The Hard Law-Soft Law Nexus: Autonomous Vehicles as a Case Study

April, 2026

The technology governance debate often focuses on the dichotomy of hard law versus soft law as competing models. Both hard law and soft law have their strengths and weaknesses. But framing soft law versus hard law as a dichotomous choice is often unrealistic—every technology will be governed by a mix of hard law and soft law. So, the more realistic questions are where hard law works best and where soft law succeeds, and how these two types of governance interact and integrate. After reviewing the strengths and weaknesses of hard law and soft law, this Article identifies six different hybrid models integrating hard law and soft law. The article then considers hybrid hard law/soft law models in the context of autonomous vehicles (AVs) as a case study. After reviewing the challenges and stunted progress of AV regulation, the analysis focuses on two recent rulemakings in which the first Trump and then Biden administrations proposed frameworks for AV regulation in the waning days of their respective administrations. Analysis of the comments submitted by the broad range of AV stakeholders shows a sharply divided community, with about half of the stakeholders adamantly supporting a hard law approach, and another half equally favoring a soft law approach. However, there was substantial support for hybrid hard law/soft law models, with all six models identified earlier in the article and two new models receiving support. Based on this analysis, the Article suggests that a hybrid hard law/soft law approach is the most promising direction for AV governance.

Introduction

Hard law and soft law are often portrayed as two conflicting and competing models for technology governance.¹ Hard law consists of traditional legal instruments that encompass enforceable statutes, regulations and treaties. Soft law is defined as instruments that provide substantive expectations, but which are not directly enforceable by government.² Examples include codes of conduct, best practices, private standards, and public-private partnership programs. Both hard law and soft law have their respective strengths and weaknesses for governing emerging technologies.³

Given the contrasting strengths and weaknesses, it should be no surprise that there are conflicting arguments on whether to use hard law or soft law to govern a particular technology or problem. But the question should not be whether to use hard law or soft law, but rather where and when to use each type of governance model. The reality is that every technology will be governed by some mix of hard law and soft law. It is analogous to the age-old debate on nature vs. nurture; we now know that the correct answer is that both nature and nurture affect every trait of a human or other living being.⁴ The proper focus is then to ask how nature and nurture interact, and how they each exercise their interdependent effects.⁵

In the same way, a more sophisticated and nuanced approach to technology governance recognizes that both hard law and soft law have an essential role to play for every technology, and we need to focus on where and when each can be most useful, including how these two broad categories of governance interact across a technology and over time. That is the objective of this Article: to evaluate how hard law and soft law may interact, using autonomous vehicles (AVs) as a case study.

Part I defines and differentiates hard law and soft law governance, including describing their respective strengths and weaknesses. Part II then considers models and examples of how hard law and soft law may interact. Part III describes the challenges and status of governing autonomous vehicles in the United States, noting the complex mix of federal, state, and local regulation, as well as existing and proposed hard law and soft law measures. Part IV provides the results of an empirical analysis of various stakeholder positions and suggestions for autonomous vehicle governance in two recent requests for comments on a framework for governing AVs undertaken by the National Highway Transportation Safety Authority (NHTSA). Many of the stakeholders expressly or implicitly addressed in their comments in those two proceedings their perspectives on the relevant roles and, in some cases, the interaction of hard law and soft law for promoting AV safety. Drawing on that analysis, Part V concludes by proposing a hybrid hard law–soft law approach for governing AV safety.

I. Hard Law Versus Soft Law

Hard law and soft governance approaches have divergent underlying premises and assumptions, and very different strengths and weaknesses. It is therefore to be expected that they are often pitted against each other, with some policymakers and scholars preferring hard law approaches, and others advocating soft law approaches. In other words, the choice is often presented as binary of hard law versus soft law.

Traditional government regulation is hard law. Legislatures or government regulatory agencies at the federal or state level adopt binding rules that are enforceable against regulated entities. Most environmental, health and safety regulations over the past half-century have been in the form of hard law rules, enforceable by government agencies.⁶ Examples of hard law rules include national ambient air quality standards enforced upon states under the Clean Air Act by the Environmental Protection Agency (EPA),⁷ occupational health and safety standards for toxic workplace exposures such as silica enforced against employers by the Occupational Safety and Health Administration (OSHA),⁸ and pharmaceutical approval requirements enforced against drug manufacturers by the Food and Drug Administration (FDA).⁹

Hard law rules have a number of advantages, which explain why they have been the presumptive form of governance in many fields. Most importantly, because they are mandatory, regulators can take enforcement measures, including imposing civil or criminal penalties, if an entity fails to comply with a hard law regulation.¹⁰ This ensures that all regulated entities are required to comply with the rule, even though compliance is never complete.¹¹ The second key benefit of hard laws is that the existence of a government regulator that oversees compliance with the regulation provides confidence to the public that the underlying problem or risks are being addressed.¹² This public confidence-generating function of government regulation is an important secondary benefit of hard law regulation.

However, hard law regulation also suffers from several limitations, the most important of which is the highly bureaucratic burden of adopting hard law rules.¹³ For example, a major U.S. regulation adopted by a federal agency in the United States must not only provide extensive evidentiary support to justify the action, but must also run the gauntlet of conducting over 20 subsidiary analyses, each of which may seem sensible on its own, but which collectively massively burden and slow the pace of rulemaking.¹⁴ The time taken to adopt a U.S. regulation has grown substantially over the past half century, especially when judicial review is involved (which it often is).¹⁵ So it has become increasingly difficult to adopt new regulations, and once a rule is in place, agencies generally lack the resources to amend and update their existing regulations on a timely basis. The combination of delays in both adopting and updating regulations results in a serious “pacing problem,” which is particularly problematic for fast-moving technologies like artificial intelligence.¹⁶

Hard law suffers from additional limitations, especially for rapidly developing emerging technologies. Regulatory agencies have limited jurisdiction, resources, staff, expertise and political will to adopt new regulations for an emerging technology such as artificial intelligence (AI).¹⁷ Regulations must be supported by strong and extensive empirical evidence, which often does not exist for a new technology, especially one as diverse and heterogeneous as AI.¹⁸ Proposed regulations are often second-guessed, watered down, or outright rejected by oversight from Congress, the Office of Information and Regulatory Affairs, or reviewing courts. Once the hard law rules are adopted, they often have poor compliance and enforcement, given scarce agency enforcement resources. Even so, hard law regulations can also block innovation by making it more expensive and burdensome for companies, especially small businesses, to create and advance new technological breakthroughs.¹⁹ And the effect of hard law rules is limited to the legal jurisdiction in which they are promulgated.²⁰

Fundamentally, hard law regulation is not the absolute fix of governance problems that it is often perceived to be. This “magic wand” perception of traditional government regulation assumes that if the government regulates a problem, the problem will be resolved completely. While properly targeted and limited government regulation can be an effective solution to a specific, well-defined problem, designing effective regulation is hard. There is a long and well-established list of problems with traditional regulation that often make regulation less effective than anticipated, or even in some cases counter-productive.²¹ For these reasons, Stephen Breyer, before being elevated to the U.S. Supreme Court, recommended that regulation be the “weapon of last resort,” used only when narrowly tailored to serious risks or societal problems.²²

In contrast to enforceable hard law rules promulgated by regulatory agencies, soft law includes many different types of instruments, including private standards, codes of conduct, principles, guidelines, ethical codes, best practices, certification programs, voluntary programs, and private-public partnerships.²³ Soft law can be promulgated by a variety of different entities, including governmental bodies, but also industry groups, individual companies, non-governmental organizations, think tanks, professional societies, standard-setting bodies, certification agencies, or any combination of the above.²⁴ Soft law addresses many of the limitations of hard law, which explains why soft law is increasingly used in many areas of governance.

A key benefit for fast-moving emerging technologies is that soft law measures can usually be enacted and revised much more quickly than hard law requirements.²⁵ Soft law programs are not required to go through the gauntlet of bureaucratic and procedural hurdles that a hard law rule must comply with,²⁶ and soft law measures are not subject to second-guessing by the courts and other administrative review requirements that often delay or sideline hard law rules.²⁷ For example, Underwriter Laboratories (UL) promulgated a private standard for AV safety in April 2020, and has already issued two updated standards in March 2022²⁸ and March 2023.²⁹ Given the potential for rapid enactment and revision, soft law is a critical part of recent proposals for more agile or adaptive governance.³⁰

Another important benefit of soft law is that it incorporates much broader expertise, resources, and perspectives in its enactment. While only government entities can enact hard law rules, any entity or group of entities can promulgate soft law measures.³¹ The government has limited supplies of expertise, experience, information, and resources, which can be improved by expanding the circle of entities that can enact soft law.³² For example, with the AV case study that is the focus of the last half of this Article, the government generally lacks the expertise and first-hand knowledge needed to effectively regulate this complex and diverse industry.³³

Other organizational aspects of soft law confer additional advantages. Hard law regulation is typically an adversarial process, in which government agencies unilaterally enact rules that they then enforce against regulated parties. In contrast, soft law is a voluntary system in which entities voluntarily agree to comply and cooperate.³⁴ Soft law is often generated through multi-stakeholder processes that include a variety of interests and perspectives, including those of industry and their civil society critics, helping to create a healthier ethos emphasizing cooperation and understanding. In addition, unlike hard law regulation in which one agency has the authority to adopt a single rule for a specific problem, a multitude of entities can put forward soft law proposals for the same problem, allowing a “1000 flowers to bloom,” and creating a competitive marketplace for the best soft law approaches.³⁵

Yet another category of benefits for soft law measures is that they are not subject to many of the restrictions that limit the effectiveness of hard law rules promulgated by federal agencies. Government agencies can only exercise the authority they have been delegated, which often omits some problems altogether, or limits the agency to only considering certain factors but not others when making regulatory decisions.³⁶ Hard law regulators are further limited by recent Supreme Court decisions preventing agencies from regulating “major questions” that have not been delegated by Congress,³⁷ and overturning the long-standing deference that courts have given to agency interpretations of ambiguous statutes.³⁸ In contrast, soft law has plenary authority, and can address any problem or consider any factor its proponents choose.³⁹ In addition, hard law rules are limited in their effect to the legal jurisdiction of the promulgating government agency. Soft law again has no such limitations, and can automatically apply across jurisdictions, making it particularly well-suited for international governance.⁴⁰

Of course, soft law has its own limitations and challenges as well. The biggest weakness of soft law is that compliance is voluntary, so there is no assurance that all relevant entities will accept conformance, and even some of those who do may comply only partially or half-heartedly, given the lack of any direct enforcement.⁴¹ There are a number of mechanisms that can provide indirect enforcement of soft law measures, such as pressure from consumer groups, trade associations, supply chain partners, insurers, liability lawyers, and the Federal Trade Commission.⁴² Nonetheless, the lack of direct enforcement mechanisms is the major limitation of soft law that adversely affects accountability. Companies can engage in “ethics washing” by purporting to adopt and follow soft law principles but in reality, are using that soft law mechanism as cover to engage in an unethical or irresponsible manner.⁴³

Related to the lack of direct enforcement, soft law measures are generally less trusted by the public than hard law.⁴⁴ Knowledge that a government agency has enacted the rules and stands ready to enforce them is perceived by the public as an independent and objective check on the relevant technology or industry.⁴⁵ This role of creating public confidence is often seen as a secondary or indirect benefit of regulation, over and above the primary role in controlling potentially dangerous conduct or products. In contrast, soft law measures are often seen as the fox guarding the henhouse, as the very companies that might be involved in creating and selling the technology at issue often play a significant role in creating and implementing soft law programs.⁴⁶ This public distrust in self-regulatory measures has been exacerbated by a “techlash” caused by recent corporate wrongdoings by companies such as Facebook, Boeing, Theranos, and Volkswagen.⁴⁷

Although lack of accountability and public confidence are the two most important weaknesses of soft law, there are other limitations as well. While soft law expands the scope of entities that can create governance mechanisms beyond government regulators, they do not provide the same right as government regulation for any stakeholder to submit comments that might influence the content of the resulting rule. Soft law programs are often developed behind closed doors, with only certain pre-selected organizations or individuals at the negotiation table. Even after enactment, many soft law programs have limited transparency about how and by whom they are implemented.⁴⁸

Soft law measures are also often written in very general and broad terms, which makes holding entities accountable for conformance with such measures difficult and subjective.⁴⁹ Finally, the flip side of the soft law benefit of “letting a 1000 flowers bloom” is that the ecosystem can be flooded with dozens or more competing soft law programs for a particular technology or risk.⁵⁰ Entities must therefore choose which subset of soft law programs to adopt, which can create controversy and confusion.

II. Hard Law and Soft Law

The previous section identified key strengths and weaknesses of hard law versus soft law. Based on those factors, some problems are best addressed by hard law, while others are best left for soft law. For example, a discrete and narrowly defined technology, which is causing known harm to human health or safety, should preferentially be regulated by hard law, assuming an existing agency has regulatory jurisdiction over that risk. On the other hand, a broad and rapidly evolving technology, with uncertain benefits, risks and trajectories, is best suited for soft law, at least in the short term.

The question of whether hard law or soft should govern is thus often overly simplistic. Most technologies should and will be governed by some mix of hard law and soft law.⁵¹ .Hence, the appropriate inquiry is what are the appropriate roles for hard law and soft law in governing a particular technology, and how should this change across problems and applications, and over time. There is growing recognition that both hard law and soft law have their appropriate roles to play.⁵² Bowman and Hodge distinguish a third category of governance called “hybrid” governance that combines both hard law and soft law, distinguishing it from “state regulation” (i.e., hard law) and “civil regulation” (i.e., soft law).⁵³ There are several different models of how hard law and soft law can interact,⁵⁴ with examples provided below.

A. Government Convenes, Rewards or Orchestrates Soft Law

Many soft law programs are created, coordinated, or rewarded by government entities, which have the potential to enact hard law rules, but choose instead to use soft law approaches in specific contexts. In many cases, the governmental entity conceives, designs, and administers the soft law program on its own. A well-known example is the Energy Star Program administered by the U.S. Environmental Protection Agency (EPA).⁵⁵ EPA describes this project as a “public-private partnership” because thousands of private companies voluntarily choose to participate in the program.⁵⁶ Energy Star is a voluntary labeling program in which EPA sets energy efficiency specifications for specific types of appliances or equipment, and then private companies that meet those specifications can label their products with the Energy Star logo.⁵⁷ Many businesses, contractors, and consumers specifically look for Energy Star products, and various types of incentives, such as utility rebates for purchase of Energy Star products, encourage the purchase of Energy Star labeled products.⁵⁸ In May 2025, it was reported that the EPA plants to shut down the Energy Star program as part of an initiative to defund climate programs that are not required by statute.⁵⁹

Government entities can also incentivize private actors to take voluntary actions by rewarding them with some benefit, such as public recognition or reduced enforcement scrutiny.⁶⁰ For example, EPA’s National Environmental Performance Track program, which ran from 2000 to 2009, identified and rewarded companies that go “above and beyond” environmental compliance to achieve superior environmental performance.⁶¹ Companies that achieved benchmarks of superior environmental performance received lower priority for routine inspections and were offered public recognition, networking opportunities, and other benefits.⁶² EPA claimed that this program achieved substantial environmental benefits that could not have been achieved with traditional environmental regulation,⁶³ although a subsequent independent empirical assessment found no net benefits from the program due to business reluctance to make meaningful changes in return for the limited benefits offered by the voluntary program.⁶⁴

Government entities can also orchestrate soft law programs by using their influence to convene the relevant stakeholders, but then leaving it to industry and other stakeholders to design and implement the program.⁶⁵ For example, the National Telecommunications & Information Administration (NTIA) within the Department of Commerce has convened a number of multi-stakeholder processes to negotiate a consensus approach to emerging technologies such as unmanned aircraft systems (or drones).⁶⁶ In the case of drones, this process resulted in a voluntary best practices document for drone privacy and governance.⁶⁷

Experimental legislation represents another approach where governments utilize soft law mechanisms to test and refine regulatory approaches. These temporary legal measures allow for the testing of new policies or legal solutions within limited geographic areas or subjects.⁶⁸ Regulatory sandboxes have emerged as a specific form of experimental legislation, particularly in the context of technological innovation.⁶⁹ These controlled regulatory environments allow for the testing of novel products or processes under the supervision of regulatory authorities.⁷⁰ Sandboxes typically involve temporary regulatory relief or customized compliance assistance for a select group of participants, enabling closer collaboration between regulators and innovators. This approach aims to reduce regulatory burdens, promote innovation, and help regulators better understand emerging technologies and business models.

The first regulatory sandbox was created by the United Kingdom government in 2015 for fintech applications.⁷¹ Since then, the regulatory sandbox concept has spread to many other jurisdictions and technologies.⁷² Most recently, the regulatory sandbox model has been applied to AI governance.⁷³ The EU AI Act requires each Member State to have at least one operational AI sandbox in place by August 2026 (although nations may share cross-border sandboxes).⁷⁴ In the U.S., President Trump’s July 2025 “America’s AI Action Plan” calls on U.S. federal agencies to establish AI regulatory sandboxes around the country, “where researchers, startups, and established enterprises can rapidly deploy and test AI tools.”⁷⁵

These examples illustrate how governments are increasingly convening or creating soft law approaches like voluntary programs or regulatory sandboxes to navigate the complex landscape of technological innovation and regulation. By orchestrating these controlled environments outside of traditional regulatory approaches, governments can foster collaboration between regulators and innovators, gather valuable insights, and potentially develop more effective and innovation-friendly regulatory frameworks.⁷⁶ The role of governmental entities in convening, incentivizing, or orchestrating soft law programs is a major contributor to soft law programs, and is also a significant evolution for governmental entities from their traditional hard law role to becoming a key player in soft law.

B. Hard Law “Backstops” Soft Law

The government can create incentives for private entities to develop and conform to a soft law program by threatening to implement a hard law program if companies do not adopt a suitable soft law program on their own.⁷⁷ An example is the Health and Safety Partnership Program (“HSPP”) implemented by the North American Insulation Manufacturers Association (NAIMA) in response to a “Priority Planning Process” document published by OSHA.⁷⁸ In that document, OSHA identified 18 priorities for regulatory action, including synthetic vitreous fiber products such as fiberglass insulation, but invited industries to come forward with a voluntary alternative to regulatory action.⁷⁹ NAIMA then negotiated the HSPP with OSHA to start in 1999, and has now been implementing this voluntary program as an alternative to hard law regulation for 25 years, achieving far more safety benefits than an OSHA rule could have accomplished.⁸⁰

A related approach is when a government agency puts forth an actual or hypothetical hard law program as a regulatory “back stop” to induce greater participation in the voluntary program.⁸¹ In other words, if the private industry fails to adequately implement the voluntary program, the government will step in with a hard law program that presumably would be more burdensome for industry, thus creating an incentive to conform to the voluntary program. An example of such a strategy was the repeated discussion of regulatory backstops in the negotiation and final text of the voluntary National Low Emission Vehicle (NLEV) program at the turn of the last century. The NLEV program was a voluntary program in which vehicle manufacturers agreed to comply with a stricter but uniform national emission standard that EPA was statutorily precluded from adopting and enforcing as a hard regulation, in exchange for eliminating the different state-by-state patchwork of emission standards that was developing.⁸² An important part of the negotiation, and a key point of leverage over the vehicle manufacturers, was that if manufacturers failed to comply with the NLEV standards, states that had previously adopted the California Low Emission Vehicle (CAL LEV) standards could immediately reinstate those programs as a ‘backstop” to encourage compliance.⁸³

In the realm of standard-essential patents (SEPs) and technology standards, antitrust law enforced by the Federal Trade Commission (FTC) serves as a hard law backstop to the soft law mechanisms of industry self-regulation through Standard-Setting Organizations (SSOs) and Fair, Reasonable, and Non-Discriminatory (FRAND) commitments. SSOs often require participants to disclose relevant patents and commit to licensing their SEPs on FRAND terms as safeguards against potential anticompetitive behavior.⁸⁴ However, when these soft law mechanisms fail to prevent anticompetitive conduct, the FTC can step in with antitrust enforcement actions.

The FTC has brought cases challenging deceptive conduct and failure to abide by FRAND licensing obligations.⁸⁵ For instance, in the Dell case, the FTC alleged that Dell’s failure to voluntarily disclose a relevant patent during a standard-setting process violated Section 5 of the FTC Act.⁸⁶ This demonstrates how antitrust law can serve as a backstop when SSO disclosure policies are insufficient or not followed. Similarly, the FTC has acted against SEP holders who seek exclusionary remedies against willing licensees, viewing such conduct as potentially violating FRAND commitments and harming competition.⁸⁷

This approach creates an incentive for industry participants to adhere to SSO policies and FRAND commitments, as failure to do so could result in more burdensome antitrust enforcement actions. The FTC’s enforcement actions serve as a “backstop to monitor and enforce against conduct in the standard-setting context that harms competition.”⁸⁸ The threat of antitrust intervention thus reinforces the effectiveness of the soft law mechanisms in the standard-setting context, much like the regulatory backstops in the NLEV program encouraged compliance with the voluntary emissions standards.

A final example of the threat of hard law backstopping soft programs is the Motion Picture Association of America (MPAA) movie rating system.⁸⁹ The MPAA adopted this voluntary movie rating scheme in order to head off plans by the federal government to regulate movie content.⁹⁰ The success of this soft law program “has kept the government off the back of the film industry.”⁹¹

C. Soft Law Evolves into Hard Law

Another potential relationship between soft law and hard law is a temporal one, in which soft law evolves into hard law. It may be that the technology and governance issues for a particular issue are too uncertain and unsettled for enacting a hard law solution initially. In these situations, soft law can provide an initial governance framework, which may be just an interim solution. The soft law experience can generate information on what works and what does not to better inform the subsequent hard law enactment. Providing a pathway for growing a soft law program into a future hard law program is one way for a soft law program to succeed,⁹² although not all soft law programs need or should evolve into hard law.

The evolutionary trajectory of soft law to hard law is shown below in Figure 1.

Figure 1: Evolutionary trajectory of soft law to hard law.⁹³

An example of this soft law to hard law transition is the regulation of fuel economy savings from improvements to the trailers of heavy-duty vehicles. The Energy Independence and Security Act of 2007 required EPA and NHTSA to adopt standards to regulate carbon emissions and fuel economy for medium- and heavy-duty commercial vehicles.⁹⁴ The first such federal regulations were adopted in 2011 and focused on more efficient engines for such vehicles. The agencies considered but did not adopt standards for the trailers for such vehicles, explaining:

Neither EPA nor NHTSA is adopting standards at this time for GHG emissions or fuel consumption, respectively, for heavy-duty commercial trailers . . . The agencies recognize that aerodynamic and tire rolling resistance improvements to trailers represent a significant opportunity to reduce fuel consumption and GHGs as evidenced, among other things, by the work of the EPA SmartWay program.⁹⁵

The agencies were uncertain which technologies or improvements would improve the fuel economy of trailers, nor did they have good data to quantify any carbon reduction and fuel economy improvements from such interventions. So, as indicated in the quote above, the agencies relied on a voluntary program called “SmartWay” to test different approaches for improving the fuel economy of heavy-duty trailers.

SmartWay is a voluntary public-private program created in 2004 by EPA in partnership with the transportation industry to promote more sustainable transport,⁹⁶ and now has almost 4000 company partners.⁹⁷ The SmartWay program tested and quantified fuel economy savings from various aerodynamic devices on the trailer such as improved tires, skirts and rear fairings.⁹⁸ This test program confirmed which equipment improved fuel economy and by how much, but also reassured drivers who had concerns about the convenience and safety of some such devices.⁹⁹ The findings from the voluntary SmartWay program demonstrating successful aerodynamic interventions lead the state of California to then adopt trailer fuel economy requirements expressly based on the SmartWay certifications in its hard law regulations.¹⁰⁰ Once the California hard law regulations validated the benefits from the SmartWay program, the EPA and NHTSA subsequently adopted similar requirements (again based on the SmartWay verified technologies) into the second generation of federal heavy-duty vehicle greenhouse gas and fuel economy regulations.¹⁰¹ So here we have an example of regulating a new technology starting with soft law, progressing then to state hard law, and finally to federal hard law, with each enactment building on the findings of the previous program.

There are other examples of soft law programs evolving into hard law regulations. For example, the National Institute of Standards and Technology (NIST) Cybersecurity Framework represents an evolving example of soft law transitioning towards hard law in the cybersecurity domain.¹⁰² Initially developed as a voluntary guideline in response to Executive Order 13636,¹⁰³ the Framework provided a flexible, risk-based approach to managing cybersecurity risk.¹⁰⁴ This soft law approach allowed organizations to adapt the framework to their specific needs and risk tolerances. The Framework’s transition towards hard law began with the Cybersecurity Enhancement Act of 2014. This legislation codified NIST’s role in developing cybersecurity frameworks, giving legislative backing to what was previously a voluntary initiative.¹⁰⁵ The Act required NIST to “identify and develop cybersecurity risk frameworks for voluntary use by critical infrastructure owners and operators.”¹⁰⁶ While still emphasizing voluntary adoption, this legislative mandate represented a significant step towards formalizing the Framework’s status.

Furthermore, the Act required federal agencies to coordinate with NIST in developing standards and guidelines,¹⁰⁷ effectively integrating the Framework’s principles into government cybersecurity policies. This legislative recognition elevated the Framework’s status from purely voluntary guidance to a Congressionally mandated approach for improving cybersecurity across critical infrastructure sectors.

The Framework continues to evolve, with NIST releasing updated versions based on stakeholder feedback and changing cybersecurity landscapes.¹⁰⁸ This ongoing development, coupled with its legislative backing, illustrates how the Framework serves as a dynamic bridge between soft law flexibility and hard law authority in the rapidly changing field of cybersecurity.

Another example of soft law evolving into hard law is greenhouse gas emissions reporting requirements. Initially, many companies voluntarily disclosed their greenhouse gas emissions through initiatives such as the Carbon Disclosure Project (CDP), a soft law approach that began in 2003.¹⁰⁹ This voluntary reporting regime allowed companies and regulators to gain experience with emissions reporting practices without the constraints of hard law, which then opened the door to mandatory hard law greenhouse gas emissions reporting requirements.

The United Kingdom provides a clear illustration of this evolutionary trajectory. Prior to 2013, approximately 50% of UK-incorporated listed companies were already voluntarily disclosing greenhouse gas emissions.¹¹⁰ This soft law approach provided valuable information on reporting practices and their effects. Building on this experience, in 2013 the UK government implemented Mandatory Carbon Reporting (MCR), requiring all UK-incorporated listed companies to report their greenhouse gas emissions in annual financial reports.¹¹¹ This transition from voluntary to mandatory reporting represents a direct evolution from soft law to hard law.

The progression did not stop there. Other jurisdictions, such as Australia, Japan, Sweden, and the state of California, followed a similar path, first mandating disclosure before moving to even stricter regulations like carbon taxation.¹¹² This pattern demonstrates how soft law cannot only evolve into hard law, but can also pave the way for increasingly stringent regulatory approaches.

This evolution from voluntary reporting to mandatory disclosure of greenhouse gas emissions, and potentially to carbon pricing mechanisms, illustrates how soft law can provide a foundation for more formal regulatory structures. The experience gained through soft law approaches informs and shapes subsequent hard law enactments, allowing for more effective and well-calibrated regulations.

D. Hard Law Incorporates Soft Law

Government hard law programs will sometimes incorporate by reference soft law standards. For example, the FDA incorporates many private industry standards in its regulatory programs. The FDA medical devices program maintains an online database of over 1500 consensus voluntary standards that are recognized and accepted by the FDA in its hard law regulatory approvals of medical devices.¹¹³ At last count, over 27,000 private consensus standards have been incorporated by reference into U.S. government regulations.¹¹⁴ States also often incorporate soft law into their hard law—for example thirty-five states regulate amusement parks by requiring them to conform to otherwise voluntary ASTM safety standards for such activities.¹¹⁵

Even the European Union’s recently adopted AI Act,¹¹⁶ often heralded as the primary example of hard law regulation of AI, applies hard law to some AI applications but soft law, such as voluntary codes of conduct and private standards for many other (in fact, most) AI applications.¹¹⁷ For example, the EU relies on a non-binding soft law Codes of Practice for regulating general-purpose AI models.¹¹⁸ For high-risk AI systems, the EU AI Act requires the EU Commission to request the European Standards Association and other EU standard setting-organizations to promulgate private standards for the management of AI risks that, after being assessed and approved by the Commission, will confer providers of high-risk AI systems with a presumption of conformity with the EU AI Act.¹¹⁹

E. Parallel Governance by Hard Law and Soft Law

Parallel governance, often referred to as co-regulation,¹²⁰ occurs when both government hard law and private soft law play a role in an integrated regulatory system.¹²¹ For example, with respect to the certification of hospitals and other health care facilities that are required to receive Medicare and Medicaid funds, the federal government (through the Centers for Medicare and Medicaid Services (CMS)) has generally delegated this task to several private entities, most significantly the Joint Committee (formerly the Joint Committee on the Accreditation of Healthcare Organizations).¹²²

There are many other examples of co-regulation where a government agency regulates part of a problem using hard law and a private entity performs other quasi-regulatory tasks as non-binding soft law. For example, the FDA regulates cosmetics but outsources safety determinations for new cosmetics to an industry group called the Cosmetics Industry Review (CIR).¹²³ Similarly, the National Highway Traffic Safety Administration (NHTSA) policy for cybersecurity of motor vehicles consists of “best practices” that mostly consist of industry voluntary standards.¹²⁴ The Federal Communications Commission (FCC) relies on “frequency coordinators” who are private actors that manage voluntary cooperation to ensure that new licenses do not interfere with the spectrum of existing users.¹²⁵ The FTC relies largely on the Better Business Bureau’s National Advertising Division (NAD) to privately resolve disputes over misleading advertising, with the FTC regulatory power serving as a back-up in the relatively few instances when the dispute is not resolved by the NAD’s process.¹²⁶

Antitrust or competition law, especially when viewed from an international perspective, features national hard law that sets regulatory requirements, supplemented by international soft law in the form of economic principles and international guidelines to operationalize the hard law and promote international coordination.¹²⁷ Under this scheme of parallel governance, both hard law and soft law have important roles to play: “Soft law, with its ability to provide interpretative guidelines and foster international cooperation, is essential in contexts of rapid innovation. Hard law remains important to ensure stability and prevent abuse, ensuring that the interests of consumers and competitors are protected in a clear and uniform manner.”¹²⁸

The Norwegian offshore oil and gas regulatory system also exemplifies hard law/soft law co-regulation in practice. The Petroleum Safety Authority Norway (PSA) establishes function-based regulations that set hard law performance goals, while the industry, through bodies like Standards Norway, develops detailed soft law technical standards (NORSOK standards) that are often referenced in or incorporated into these regulations.¹²⁹ This approach integrates government hard law with industry-developed soft law to create a comprehensive regulatory framework that balances flexibility with accountability.¹³⁰

Multiple instances of co-regulation exist within this system. The PSA mandates risk assessment and management, but the specific methodologies and best practices are often developed by industry groups and technical committees.¹³¹ This allows for rapid adaptation to technological advancements and emerging risks. The Working Together for Safety (SfS) program, an industry initiative, develops safety recommendations that, while voluntary, are widely adopted and often inform regulatory expectations.¹³² This program demonstrates how industry self-regulation can complement formal regulatory requirements. This co-regulatory approach allows for dynamic adaptation to technological changes and emerging risks, while maintaining strong government oversight of the high-risk offshore environment.¹³³

F. Hard Law Enforcement of Soft Law Non-Compliance

A final model of how soft law and hard law may interact is one in which the government uses hard law to enforce a private actor’s non-conformance with its soft law commitments. While a soft law measure is, by definition, not usually enforceable by government, a back-door indirect enforcement might exist where an actor’s failure to conform to its public commitment to a soft law measure rises to the level of “unfair” or “deceptive” practices. The FTC, under its Section 5 general authority to take enforcement actions against deceptive and unfair business practices,¹³⁴ could take action against a company that publicly commits to comply with a certain voluntary code of conduct or best practices but then fails to live up to its commitment.¹³⁵

The FTC has already used its power against companies that fail to comply with their voluntarily adopted privacy policies.¹³⁶ For example, FTC took enforcement action against a company that falsely claimed it adhered to the guidelines of an FTC-approved safe harbor program pursuant to the Children’s Online Privacy Protection Act (COPPA), such as the Better Business Bureau’s Children’s Advertising Review Unit (CARU) program.¹³⁷ Similarly, the FTC entered into a consent decree with twelve companies who falsely claimed that they were abiding by the U.S.-EU Safe Harbor, a voluntary international privacy framework.¹³⁸ The consent decree prohibited the twelve companies “from misrepresenting the extent to which they participate in any privacy or data security program sponsored by the government or any other self-regulatory or standard-setting organization.”¹³⁹ Google was also hit with a $22.5 million penalty for failing to comply with its voluntary commitment to conform to an industry standard (the Self-Regulatory Code of Conduct of the Network Advertising Initiative).¹⁴⁰

III. Autonomous Vehicle Governance

Shifting gears, this section introduces the regulatory challenges of autonomous vehicles (AVs), which will be used as a case study of hybrid hard law/soft law governance. AVs are vehicles that can operate without a human driver. This capacity exists in a wide range of technological capabilities, from adaptive cruise control to vehicles with no steering wheels or gas pedals that cannot be operated by a human at all. The biggest issue with AVs is safety—it is both the technology’s greatest potential benefit and its greatest concern.¹⁴¹ The primary legal issue presented by AVs is how to ensure that AVs are adequately safe.

Several companies are currently developing and testing AVs in the United States. Waymo, a Google (Alphabet) company, is the industry leader in developing AVs, even though it is not a traditional vehicle manufacturer. Waymo has operated its autonomous vehicles without a human safety driver on over 60 million miles of U.S. public roads (as well as more than 20 billion miles of simulation driving).¹⁴² Data published by both Waymo and the insurer Swiss Re demonstrate that the Waymo AV is 70-90 percent safer than cars driven by humans in similar environments.¹⁴³

Waymo started carrying passengers in its AVs (initially with a safety driver in the driver’s seat) in Arizona in 2017. It expanded this testing to its Waymo One ride-hailing program, which began operating with no safety driver starting in 2020.¹⁴⁴ It has now expanded its commercial ride-hailing system without safety drivers to not only a much larger portion of the Phoenix area, but also to San Francisco, Los Angeles, Austin, and Atlanta, with plans to soon start service in many other cities, including Miami, Washington, D.C., Philadelphia, and New York City.¹⁴⁵ Tesla has also started operating autonomous “robotaxis” in Austin in June 2025 and is planning to expand this service to other cities soon.¹⁴⁶ A number of other companies, including Volkswagen, Zoox (owned by Amazon), and Avride, are also testing AVs on public streets in various U.S. cities.¹⁴⁷

Under the U.S. federal system, regulatory responsibility for motor vehicles is shared between the federal government and state and local governments. The federal government has the sole responsibility and authority to specify the technological requirements for new motor vehicles. State and local governments have jurisdiction over in-use performance such as licensing of vehicles, licensing of drivers, operation of vehicles, and traffic control. This traditional demarcation is challenged by autonomous vehicles, since the vehicle is essentially the “driver.” ¹⁴⁸ This thus blurs the line between federal regulation of the vehicle and state regulation of the driver.¹⁴⁹ Currently, both the federal and state/local governments are trying to straddle this blurry jurisdictional line with their respective and incomplete regulatory activities. The result is a regulatory mess and stalemate—”Nobody—not federal lawmakers, not state lawmakers, and not regulatory agencies—seems super clear on where they are going, and the journey has been marked by dead ends, turnarounds, and uncertainty.”¹⁵⁰

The federal government has the primary responsibility for approving AVs, and so this discussion will focus on federal governance of AVs.¹⁵¹ Two key regulatory entities exist at the federal level for autonomous vehicles: the U.S. Congress and the National Highway Traffic and Safety Administration (NHTSA). Congress must pass new legislation if it wants to delegate regulatory authority beyond what exists for traditional vehicles, but it has yet to legislate on AVs. In the absence of any such Congressional action, NHTSA, which has the delegated authority to regulate the safety of new motor vehicles, is limited to exercising the limited authority it previously had for non-autonomous vehicles.

A. Congress

Despite numerous attempts to enact new legislation to govern autonomous vehicles, Congress has yet to pass any major legislation relating specifically to autonomous vehicles. Congress came closest to passing comprehensive automated vehicle legislation in 2017 in the 115^th Congress. The House passed by unanimous voice vote a bill (H.R. 3388), known as the SELF DRIVE Act, that would have applied to several aspects of automated vehicles, including preemption of state laws that would regulate the design or performance of autonomous vehicles, a requirement for NHTSA to require manufacturers of autonomous vehicles to submit a “safety assessment certification,” and requiring cybersecurity plans for autonomous vehicle models.¹⁵² The legislation would also have expanded the number of exemptions that NHTSA can grant from compliance with FMVSSs from 2500 per manufacturer per year to up to 100,000 vehicles per manufacturer per year.¹⁵³

A similar but somewhat different bill was introduced in the Senate (S. 1885), called the AV START Act,¹⁵⁴ but the bill did not reach the floor after several Senators raised concerns about the proposed preemption of state laws and the large number of exemptions from FMVSSs that were included in the proposed Senate bill (and adopted House bill).¹⁵⁵ Both the House and Senate bills would preempt state laws regulating the design or performance of autonomous vehicles, and both bills supported the traditional role of states in regulating vehicle registration, driver licensing, driving education, insurance, law enforcement, and accident investigation. However, given the complexity of separating these traditional functions from the design of autonomous vehicles, consensus could not be reached on the appropriate scope and language for preemption, and as a result no legislation was adopted.¹⁵⁶

Although discussions and hearings on proposed autonomous vehicle legislation have continued in Congress since the 2017 close call, no bills have advanced in subsequent sessions. While all stakeholders claim to agree on the need for new Congressional legislation for autonomous cars, no agreement exists on what that legislation should provide, and therefore, the legislative stalemate continues to this date. In mid-2022, two prominent members of the House of Representatives, one Democrat (Debbie Dingell) and one Republican Bob Latta), launched a bipartisan Congressional Autonomous Vehicle Caucus.¹⁵⁷ The objective of the bipartisan caucus is to educate and mobilize members of Congress and their staff “to work towards a national framework so that AVs can continue to safely expand throughout the country.”¹⁵⁸ Co-chair Latta told media sources that he hoped that Congress could enact legislation providing a national framework for autonomous vehicles in the 2023-2024 time period.¹⁵⁹ Yet, no real progress has been made at the Congressional level for regulating AVs, demonstrating the pacing problem in action.¹⁶⁰

B. NHTSA

NHTSA is a federal agency within the U.S. Department of Transportation (DOT) that has been delegated regulatory authority by Congress to oversee various aspects of new motor vehicles, including safety and fuel economy. Safety is the primary regulatory issue for autonomous vehicles, so not surprisingly, NHTSA has been the most active U.S. federal agency on autonomous vehicle safety issues. This activity has included (a) some limited actions regarding Federal Motor Vehicle Safety Standards (FMVSSs), (b) annual reports providing policy guidance for autonomous and connected vehicle development, (c) various transparency and reporting measures to increase the knowledge of industry, policymakers, and the public on autonomous vehicle developments, and (d) accident investigations and statements. These four areas of NHTSA action are summarized below.

FMVSSs

NHTSA has been given two primary regulatory tools to enforce motor vehicle safety, which apply to non-autonomous and autonomous cars alike. First, NHTSA can promulgate FMVSSs, to which all new motor vehicles sold or imported into the United States must be certified to comply. Second, NHTSA can also recall vehicle models for safety defects when the agency determines that a vehicle model is unsafe in operation, even if the vehicle complies with all existing FMVSSs.¹⁶¹ The rise of AVs has created two issues: (i) whether new or updated FMVSSs are needed for AVs, and (ii) whether AVs should be exempted from some existing FMVSSs.¹⁶²

There are currently some 73 FMVSSs (codified at 49 CFR Part 571) that specify performance requirements for crash avoidance and crashworthiness in various types of accidents, as well as requirements for specific components of a vehicle, such as controls and displays, braking systems, windshield wipers, airbags, seat belts, and mirrors. For example, FMVSSs mandate equipment such as steering wheels and brake pedals in vehicles that may no longer be relevant for advanced AVs.

AVs are required to comply with the same FMVSSs that conventional vehicles are required to meet, unless NHTSA exempts them or revises a FMVSS for autonomous vehicles. As noted above, NHTSA can currently provide up to 2500 exemptions from FMVSSs per manufacturer per year, and one of the major issues in U.S. autonomous vehicle policy is whether the number of permissible exemptions should be increased for AVs.

Under President Biden, NHTSA completed its first significant FMVSS rulemaking with regard to AVs, modifying the FMVSS for occupant protection (crashworthiness) to accommodate AVs.¹⁶³ The revised standard permits vehicles to operate without a steering wheel or a “driver’s seat” for advanced AVs, provided they still meet the same level of safety for any vehicle occupants.¹⁶⁴

NHTSA Policy Guidance

NHTSA has generally been supportive of developing AVs, with its parent agency, the DOT, noting the potential of this technology to “lead to breakthrough gains in transportation safety.”¹⁶⁵ NHTSA’s evolving policy towards autonomous vehicles was initially documented in a series of annual reports that began under the Obama administration in 2016 and were extended by four reports by the first Trump Administration over the period from 2017 to 2021.

a. Obama Administration

The first NHTSA policy report entitled “Federal Automated Vehicles Policy” was issued in 2016 under the Obama Administration and set the framework for NHTSA’s AV policy in a report that is over 100 pages long.¹⁶⁶ In that document, NHTSA adopted the SAE nomenclature of Level 0 to Level 5 autonomous vehicles based on the level of autonomy, with Level 0 having no autonomy and Level 5 having full autonomy. NHTSA uses the term “Highly Automated Vehicles” to refer to vehicles in Levels 3 through 5.

The document also identified 15 practices and procedures that it expected AVs and their manufacturers to meet, including recommendations relating to data recording, privacy, crashworthiness, system safety, compliance with applicable federal, state and local laws governing vehicle operation, and object and event detection and response. Manufacturers were advised to issue voluntary Safety Assessment Letters that set forth their conformity with these requirements. This recommendation has now developed into Voluntary Safety Self-Assessments that most manufacturers have posted on the NHTSA website (discussed below).

The Obama 2016 report also set forth a model state policy for autonomous vehicles. It included recommendations for states to designate a lead agency responsible for AV testing in the state, to create an application process for manufacturers to test AVs in the state, and to make changes to motor vehicle registration, liability, and insurance requirements. The 2016 policy did not call for new regulations for AVs, although it did raise the possibility of expanding the number of exemptions that NHTSA may grant autonomous vehicles for FMVSSs, and raised the possibility of a new statutory framework that Congress might enact.

b. First Trump Administration

Donald Trump was first elected U.S. President from 2016 to 2020, and his administration had a radically different approach to most regulatory issues than his predecessor. Yet the NHTSA policy on AVs under President Trump represented an evolution or progression rather than a sharp departure from the previous NHTSA policy under President Obama. The DOT/NHTSA under President Trump published three update reports to the initial Obama 2016 framework approximately one year apart in 2017 (“Automated Vehicles 2.0”),¹⁶⁷ 2018 (“Automated Vehicles 3.0”),¹⁶⁸ and 2020 (Automated Vehicles 4.0”),¹⁶⁹ followed by a “Automated Vehicles Comprehensive Plan” issued in the final days of the Trump Administration in January 2021.¹⁷⁰ While emphasizing the central importance of safety, these reports called for a voluntary approach to autonomous vehicle safety that relied primarily on industry standards rather than new government regulations.¹⁷¹ The Trump-era reports also called on manufacturers to share with NHTSA and make public voluntary safety self-assessments.

To promote transparency, NHTSA provides information on AV testing and development by encouraging manufacturers to prepare and post Voluntary Safety Self-Assessments (VSSAs), which provide information on each company’s self-driving testing programs and approach to safety.¹⁷² As of this writing, 27 companies have posted their VSSA on the NHTSA website,¹⁷³ including every AV company in the U.S. except for Tesla. In addition, NHTSA launched its Automated Vehicle Transparency and Engagement for Safe Testing Initiative, known by its short-hand of the AV TEST Initiative, in June 2020, which is a partnership with states, local governments, and AV manufacturers.¹⁷⁴ The goal of this Initiative is to provide the public with information on AV testing organized by manufacturer, location, and road type, through an online database tool.¹⁷⁵

c. Biden Administration

The trend of annual AV policy reports came to an end with the Biden administration, which took office in January 2021. However, Secretary of Transportation Pete Buttigieg testified to Congress shortly after the new administration took office that the administration’s priority would be to “establish safety and establish certainty” for the AV industry, and called upon Congress to work with the administration to create a new regulatory framework for such vehicles.¹⁷⁶ The Biden administration also focused on disclosure from AV manufacturers (discussed below) and voluntary recalls whenever a problem was identified with a particular AV model.

During the Biden administration, NHTSA took an active role in the reporting and investigation of accidents involving autonomous and semi-autonomous vehicles. On June 29, 2021, NHTSA issued a “Standing General Order” requiring “manufacturers and operators of vehicles equipped with SAE Level 2 advanced driver assistance systems (ADAS) or SAE Levels 3-5 automated driving systems (ADS) to report crashes.”¹⁷⁷ The stated purpose of the Order is to “help the agency identify potential safety issues and impacts resulting from the operation of advanced technologies on public roads and increase transparency.”¹⁷⁸

The Order, which was amended on August 5, 2021 to provide greater clarity,¹⁷⁹ requires manufacturers of Level 2 ADAS or Levels 3-5 ADS-equipped vehicles to report within one day of learning of a crash involving their vehicle that results in a hospital-treated injury, a fatality, a vehicle tow-away, or an air bag deployment, or which involves a vulnerable road user such as a pedestrian or bicyclist. The manufacturer must submit an updated report within 10 days after learning of the crash. In addition, manufacturers must report all other crashes involving an ADS-equipped vehicle that involve an injury or property damage, and these reports must be updated monthly with new or additional information. The Standing General Order provides a specific form to standardize the accident reports.

NHTSA used this accident reporting data as a precursor to regulatory enforcement of AV safety under its authority to order recall of any vehicle model that it concludes has a “defect” that presents an unreasonable risk to public safety, even if the vehicle complies with all FMVSSs.¹⁸⁰ NHTSA has used this authority several times to order safety recalls of AV models that exhibit aberrant behavior.¹⁸¹

d. Second Trump Administration

In the first few months of the second Trump administration, NHTSA and DOT indicated they would be rolling out a new Automated Vehicle Framework that would be part of DOT’s “Innovation Agenda.”¹⁸² The AV Framework is based on three principles: (i) prioritize safety; (ii) unleash innovation; and (iii) enable commercial deployment.¹⁸³ To date, the practical impacts of this new Framework are marginal, and it includes continuing but streamlining the mandatory AV crash reporting under the General Order¹⁸⁴ and expediting exemptions by AVs from FMVSSs.¹⁸⁵

3. Private Standards for AV Safety

While Congress and NHTSA have been spinning their tires in trying to find ways to regulate AVs, private standard setting bodies have quietly and quickly moved to put into place a network of private standards.¹⁸⁶ The U.S. Department of Transportation recognized the value of these private standards in 2018:

Voluntary standards offer flexibility and responsiveness to the rapid pace of innovation, can encourage investment and bring cost-effective innovation to the market more quickly, and may be validated by private sector conformity assessment and testing protocols . . .¹⁸⁷

These private standards cover a variety of subjects. One of the first private standards for AVs was one promulgated by the Society of Automotive Engineers (SAE) that defines relevant stages or levels of vehicle automation.¹⁸⁸ NHTSA and most other AV policymakers generally follow this SAE classification of AVs.¹⁸⁹

A network of other private standards has been developed to promote different aspects of AV safety by a variety of different standard-setting organizations.¹⁹⁰ These AV safety standards include ISO 21448 entitled “Safety of the Intended Functionality” (SOTIF);¹⁹¹ ANSI/UL4600 “Standard for Safety for the Evaluation of Autonomous Products;”¹⁹² IEEE P2846 standard on “Assumptions for Models in Safety-Related Automated Vehicle Behavior;”¹⁹³ and SAE 3237standard for “Driving Assessment (DA) Metrics for Automated Driving Systems.”¹⁹⁴ Several other private standards are being developed to supplement these initial AV safety standards.¹⁹⁵

Although all these standards govern AV safety, they have limited overlap and duplication, but rather address different aspects of AV safety and fit together like pieces of a jigsaw puzzle.¹⁹⁶ These private standards provide many of the benefits of soft law measures generally described above.¹⁹⁷ For example, private standards can be adopted and revised much more quickly than traditional regulation—private standards such as UL 4600 have been adopted and revised two times within a five-year period,¹⁹⁸ whereas NHTSA has yet to adopt any hard law regulations for AVs after over a decade of deliberations. Another advantage of soft law is that it utilizes the industry experts who are actually developing and testing a technology, and who thus have first-hand knowledge of the rapidly developing technology that the government usually lacks.¹⁹⁹

IV. NHTSA Frameworks for Automated Driving System Safety

Perhaps the most far-reaching NHTSA initiatives during both the first Trump Administration and the Biden administration were the publication of documents discussing an overall framework for AV governance for public comment. In the first Trump Administration, NHTSA published on December 3, 2020, almost a month after President Trump had lost re-election, an advance notice of proposed rulemaking (ANPRM) in the Federal Register laying out and asking for public comment on an initial framework for automated driving system safety.²⁰⁰ NHSTA explained that it is “considering the creation of a governmental safety framework specifically tailored to ADS [Automated Driving Systems],” and was seeking comment on the selection and design of “the structure and key elements of a framework.”²⁰¹ It specifically called for comment on “appropriate administrative mechanisms to achieve the goals of improving safety, mitigating risk, and enabling the development and introduction of new safety innovations,” and then provided 25 specific questions on which it invited comments.²⁰²

NHTSA’s 2020 ANPRM identified a number of challenges to adopting FMVSSs for AV in the near future, suggesting that this was something it was not considering at this time but may return to this option in the future. Factors identified by NHTSA included the fact that autonomous vehicle system development was still in their early stages of development, the agency lacks the empirical data on safety and feasibility required to set a FMVSS, that premature adoption of FMVSSs could distort manufacturer plans and actually jeopardize vehicle safety, and that FMVSSs typically apply to all applications of a vehicle, not to specific operational design domains like autonomous vehicles. Instead, NHTSA suggested that in “the coming years, NHTSA will address unreasonable safety risks that arise using its defect investigation and remediation authority.”²⁰³ In addition, NHTSA discussed reliance of various reporting and information communication provisions, and private standards to set a voluntary safety framework for autonomous vehicles. The key factor that NHTSA requested comment on was whether the agency should adopt enforceable rules on autonomous vehicle safety, likely in the form of new or amended FMVSSs, or instead should rely on a cooperative, voluntary approach.

Just over four years later, the outgoing Biden administration published its own 60-page Federal Register notice in the final days of its tenure, this time framed as a proposed rule.²⁰⁴ The proposal was named the ADS-equipped Vehicle Safety, Transparency, and Evaluation Program (AV STEP).²⁰⁵ In the proposal, NHTSA acknowledges it does not have sufficient evidence to adopt FMVSSs for AVs – “the data, methods, and metrics to support such standards do not yet exist.”²⁰⁶ So NHTSA instead proposed the AV STEP initiative as a “voluntary program” in which AV manufacturers would have “an opportunity to demonstrate their operational safety and their commitment to transparency for their vehicles and operations by engaging in a national program with well-defined participation and reporting criteria focused on safety.”²⁰⁷

Under the proposed program, a manufacturer could apply to participate in the program by submitting to NHTSA “information and data related to the safety of the design, development, and operations of ADS-equipped vehicles for their intended deployment under the program.”²⁰⁸ NHTSA would then supervise an independent third-party review of the safety information for the manufacturer’s vehicle and could ask questions or request additional data.²⁰⁹ The primary focus of this independent assessment would be the manufacturer’s conformance with applicable industry standards that are relevant to the vehicle being assessed.²¹⁰

Once NHTSA was satisfied with the applicant’s data submission, the manufacturer would be accepted into the program, which would “reflect a determination by NHTSA that the applicant has provided evidence showing it followed well-documented engineering processes and has the needed technical, operational and management resources in place to mitigate safety concerns.”²¹¹ However, “[a]cceptance into the program would not be an assurance of safety, a validation of the ADS technology, or a guarantee that the applicant will execute its operational oversight functions as described.”²¹² A manufacturer who is accepted into the program would be obligated to submit periodic performance reports as well as reports triggered by any “event.”²¹³

The Trump and Biden Federal Register notices on a governance framework for AVs provide a useful gauge of stakeholder views on hard law versus soft law governance of AVs. Both notices were issued after the incumbent political party in the White House had lost re-election, so there was little chance that these lame-duck proposals would be implemented by the subsequent administration. Nonetheless, both notices squarely raised the question of whether AVs should be governed by hard law or soft law (as discussed above), and so the public comments filed in response to the notices provide insight on the positions of the key stakeholders on hard law versus soft law, and whether they propose or support any hybrid hard law/soft law strategies.

A. The 2020 Trump NHTSA ANPRM

NHTSA received 750 comments in response to its November 2020 request for comments, presenting a diverse array of positions and suggestions.²¹⁴ A key issue on which commentators were sharply divided was whether NHTSA should rely primarily on private standards promulgated by various standard-setting bodies and other soft law provisions, or should adopt and enforce its own binding hard law regulations such as new AV-specific FMVSSs.²¹⁵

The analysis below examines the comments received regarding soft law and hard law. Most comments submitted were from individuals, usually in the form of a short form letter, providing no original substantive analysis or recommendations. We excluded these individual comments from our analysis, both because they were usually without substance and because we have no way to determine how representative their opinions were of the general population. Instead, we have focused our analysis on the 125 comments submitted by stakeholder organizations. A broad array of organizations submitted substantive comments, representing the range of stakeholders with an interest in autonomous vehicle development and regulation. This set of comments therefore provides a useful resource to get a snapshot of the autonomous vehicle ecosystem with respect to hard law and soft law preferences.

The analysis below first categorizes each organizational comment by whether that entity favors a hard law approach, a soft law approach, or a combination of hard law and soft law. We present cumulative preferences for all organizations and then disaggregate by organizational category or sector. Then, we provide examples of comments favoring hard law, soft law or a hybrid approach.

1. Data on Hard Law vs Soft Law Preferences

The Table below organizes the significant comments to NHTSA on its 2020 Self Driving Framework ANPRM by the type of organization and the organization’s recommendation on whether NHTSA should pursue a hard law, soft law, or hybrid soft law-hard law approach. There was substantial disagreement on the preferred approach—31% overall favored a hard law approach, 44% favored a soft law approach, and 25% recommended a hybrid hard-soft law approach. The most lop-sided sector was industry, with 57% of industry commentators favoring a soft law approach, even though 16% of industry favored a hard law approach.²¹⁶

Table 1: Summary of Comments on Trump Nov. 2020 NPRM

Type of Commentator	Number of Commentators	Favoring Hard Law	Favoring Soft Law	Favoring Hybrid Hard-Soft Law
Industry	49	8 (16%)	28 (57%)	13 (27%)
NGO	27	13 (48%)	9 (33%)	5 (19%)
Professional	19	5 (26%)	9 (47%)	5 (26%)
Academic	21	9 (43%)	8 (38%)	4 (19%)
Governmental	9	4 (44%)	1 (11%)	4 (44%)
Total	125	39 (31%)	55 (44%)	31 (25%)

2. Comments Favoring Hard Law Approach

Just under a third (39/125) of the comments submitted on the 2020 ANPRM favored a pure hard law approach. These hard law positions came from a variety of perspectives and stakeholders. Unsurprisingly, civil society organizations advocating for vehicle safety were most clear and adamant in their advocacy for AV hard law rules. For example, Advocates for Highway and Auto Safety emphasized the need for robust and mandatory federal oversight and safety standards, rather than relying on voluntary industry measures.²¹⁷ It recommends that NHTSA implement specific testing safeguards for AVs “since voluntary approaches have proven to be inadequate.”²¹⁸ The proposed mandatory regulatory measures would aim to address critical safety concerns and ensure thorough evaluation of AV systems before they end up on public roads.²¹⁹ They advocate for the creation of FMVSSs especially tailored to AVs to address shortcomings in current AV technology.²²⁰

The Center for Auto Safety, another civil society organization advocating for safer vehicles, was harshly critical of NHTA’s historical reliance on voluntary measures to govern AVs and called for comprehensive hard law regulation.²²¹ “Instead of additional non-binding guidance,” NHTSA should “establish minimum overall safety performance requirements for ADS so that both industry and auditors will have an appropriate reference base for safety evaluation.”²²²

The American Association of State Highway and Transportation Officials (AASHTO), representing state transportation regulators, argued that “regulation is needed now when it could lead to more widespread testing and greater public trust in the technology.”²²³ The organization called for “[f]ederal minimum safety performance requirements for testing, piloting, or running limited deployments of ADS-equipped vehicles on public roads” that should be implemented in new FMVSSs for AVs.²²⁴ According to the Association, “[e]xisting consensus standards alone are inadequate to define the safe operating envelope of the many types of proposed ADS.”²²⁵

Several unions filed comments strongly supporting a hard law approach. For example the Transit Workers Union field comments advocating for the immediate imposition of strict enforceable rules for AV safety, arguing that “[t]he Trump DOT’s hands-off approach to AV regulation—and its embrace of industry-led standard-setting and loose agency guidelines—was dangerous and exposed the public to potentially severe safety risks.”²²⁶ Similarly, the Transportation Trades Department, AFL-CIO (TTD) advocated “a comprehensive framework for ADS safety” that is based on “true regulatory oversight” rather than “non-binding guidance and voluntary consensus standards [that] are wholly inadequate to ensure safety and security in transportation.”²²⁷

MITRE, a non-profit research company, also supported a hard law regulatory approach that implements a mandatory, three-tiered Safety Management System (SMS) for AVs.²²⁸ This system would cover safety across design, testing, and post-release phases, ensuring that AV technologies meet safety standards at each stage.²²⁹ The design risk management phase requires developers to apply risk management protocols to functional safety models to mitigate technical debt.²³⁰ In the Testing Risk Management phase, developers must show that the prototype performs consistently with the design model.²³¹ Such testing should verify that data systems align with NHTSA standards. Finally, the release risk management phase is grounded in continuous validation.²³² It requires ongoing monitoring of key safety metrics to confirm that AVs maintain safe operation amidst evolving conditions. A lapse in reporting may have serious consequences. By mandating this structure, MITRE aims to enforce consistent safety practices throughout the lifecycle of AV development.²³³

The insurer Munich Re submited comments supporting a national safety framework for AVs and for NHTSA to “require a clear indication by the manufacturer that a particular vehicle is capable of performing sensing, perception, planning, and control at a level of proficiency that is in alignment with the proposed safety framework.”²³⁴ Munich Re also proposed a requirement for AV developers to define and publicly disclose each vehicle’s Operational Design Domain (ODD).²³⁵ This would legally mandate developers to clarify the specific conditions under which an ADS can safely operate.

A final example of a commentator that supported a hard law approach is the National Transportation Safety Board (NTSB), which advocated for mandatory implementation of minimum performance standards and AV data reporting requirements.²³⁶ The NTSB also called for mandatory incorporation of more robust collision avoidance test procedures into the New Car Assessment Program for new vehicles, which will help save lives and would go a long way in building public confidence in the capabilities of automated technologies.²³⁷

3. Comments Favoring a Soft Law Approach

A purely soft law approach was the most common response to the 2020 NPRM, with 55 organizations representing 44 percent of organizational responses supporting a pure soft law governance framework. Some examples of such comments are summarized below.

The Alliance for Automotive Innovation, the major trade association representing vehicle manufacturers, supported voluntary approaches at this time, while recognizing that hard regulation may be appropriate at some time in the future if and when supported by sufficient evidence.²³⁸ The Alliance endorsed “the Agency’s current approach of providing guidance, encouraging transparency through voluntary reporting, and updating existing Federal Motor Vehicle Safety Standards (“FMVSS”) to remove barriers that inhibit the introduction of this promising technology.”²³⁹ The Alliance again agreed with NHTSA statements that “meaningful data is not yet available about the on-road experience of these systems that can be analyzed to determine whether a safety need exists to justify a new FMVSS rulemaking.”²⁴⁰ Many individual manufacturer members of the Alliance filed their own separate comments endorsing the Alliance’s soft law approach.²⁴¹

The Self-Driving Coalition for Safer Streets, representing over a dozen companies developing AVs, advocated “voluntary mechanisms that can be used to implement the elements of a safety framework, including for monitoring, influencing, and encouraging greater care.”²⁴² It recommended that “NHTSA should avoid at this time issuing traditional FMVSS for ADS that are focused on discrete safety features with prescriptive metrics and test procedures.”²⁴³ Instead, “NHTSA in the near term should prioritize developing and administering an ADS safety framework that is flexible and adaptive, and that employs tools such as guidance documents and consumer materials that will provide the information needed for the safe deployment of ADS technology without imposing prescriptive rules that may prove unsuitable when ADS deployment reaches a more mature stage.”²⁴⁴

CTIA—the Wireless Association, filed comments supporting “NHTSA’s expressed intent to take a measured, light touch approach as it considers a new regulatory framework for assessing the safety of automated driving systems, which are still in the research, development and testing phase.”²⁴⁵ Given that the AV technology is “nascent,” CTIA argues that “adopting rigid rules prematurely could have unintended consequences both on consumer safety as well as vehicle technology development.”²⁴⁶ The CTIA therefore encouraged “NHTSA to design the safety framework to allow for more industry-led and self-regulatory mechanisms in this still emerging state in the field.”²⁴⁷

The trucking AV company Locomation (which subsequently shut down) submitted comments backing a voluntary self-certification approach based on UL 4600, enabling AV developers to create safety cases suited to their specific needs.²⁴⁸ This model offers flexibility, allowing developers to adapt safety practices to different environments while maintaining high standards, all without formal regulatory oversight. Initial reliance on such “industry-developed, consensus-based safety standards is a smart and logical foundation for future regulations.”²⁴⁹

AV manufacturer Pony.ai advocated for a technology-neutral, performance-focused voluntary framework, which emphasizes the desired safety outcomes rather than prescribing specific technologies.²⁵⁰ This approach would allow ADS developers to use a variety of solutions to meet safety goals, fostering innovation while ensuring that all systems achieve the same level of safety. It also ensures that any proposed regulation or standard provides “reasonable notice” and “ensures due process.”²⁵¹

4. Comments Favoring a Hybrid Hard Law/Soft Law Approach

A significant percentage of each sector (25% overall, ranging from 19-44% by sector) supported a hybrid hard law/soft law approach. A total of thirty-one comments, from every type of organization, endorsed a hybrid approach to AV governance. Many different alternatives for hybrid hard law-soft law governance were suggested, and these recommendations covered the six various forms of hard law-soft law interactions described earlier in this paper.²⁵²

The first model for hard law/soft law interaction is for the government to convene or orchestrate soft law approaches.²⁵³ The Society of Automotive Engineers (SAE), which promulgates voluntary industry standards for AVs, submitted comments urging NHTSA to consider and give emphasis to the private standards adopted by SAE and other standard-setting organizations.²⁵⁴ However, SAE recommended that this interaction be two-way. SAE invited “input from NHTSA regarding priorities, needs, and gap analysis would be valuable to SAE committees as they develop their priorities so that published open consensus standards can continue to inform NHTSA’s efforts and, as appropriate, be incorporated in US DOT policy and regulations expeditiously.”²⁵⁵ In other words, the industry group is inviting the hard-law regulator (NHTSA) to provide guidance on future soft law measures.

The American Association of State Highway and Transportation Officials (AASHTO) urges NHTSA to “take a more proactive role in supporting and establishing ongoing collaboration opportunities.”²⁵⁶ Specifically, AASHTO recommends that NHTSA play the role of “convenor” to “bring together state and local DOT representatives, research partners, USDOT, auto industry, original equipment manufacturers, and technology vendors” to develop AI governance frameworks.²⁵⁷ Aurora recommended a similar convening role for NHTSA to coordinate soft law and hard law approaches to AV safety by creating an “Industry-Government Collaborative Roadmap” for AV development and oversight.²⁵⁸

The second model of hybrid hard/soft law is for hard law to backstop soft law.²⁵⁹ NVIDIA incorporated this model in its comments on the 2020 ANPRM.²⁶⁰ NVIDIA proposed a framework in which “NHTSA could release early draft proposed standards in the form of voluntary guidance open to feedback by industry experts, including those who have already contributed to international consensus safety standards.”²⁶¹ If the industry responded responsibly, then, “[i]deally, NHTSA should not need to issue its own standards but should identify which industry standards are expected to be adhered to by the manufacturers.”²⁶² However, if the industry response was not satisfactory, then NHTSA would presumably proceed with adopting its own standards.

The third model is for soft law to evolve into hard law.²⁶³ Kodiak Robotics supported this dynamic model of soft law transitioning to hard law in its comments.²⁶⁴ For the present time, the company recommended that “rules-based guidance is most appropriate to ensure that ADS developers incorporate safety best practices and ensure that NHTSA builds public confidence in ADS technology.”²⁶⁵ It specifically advocated for reliance on soft law guidance: “[ru]les-based functional safety processes such as ISO 26262 have long been critical to safety assessment in the automotive industry. They have a long record of success . . . “²⁶⁶ As the technology develops, however, traditional regulatory “performance standards may become appropriate and necessary.”²⁶⁷ In summary, “in the long run, thoughtful regulation will be critical to assuring public acceptance of ADS-equipped vehicles. At the same time, premature or unnecessarily burdensome regulations could prove to be a limit on innovation, needlessly holding back ADS technology.”²⁶⁸

Model four is for hard law to incorporate soft law.²⁶⁹ The comments of Underwriters Laboratories (UL), which among others promulgates voluntary consensus standards for technologies such as AVs, supported this model.²⁷⁰ UL urged NHTSA to “leverage public-private partnerships in developing public policy by incorporating consensus-based standards, available accreditation schemes, and globally recognized practices to meet its compliance interests.”²⁷¹ It specifically called for the NHTSA regulatory safety framework to require “[c]onformance with an agreed set of appropriate industry safety standards addressing important aspects of ADS development such as ISO 26262 (Functional Safety), ISO 21488 (SOTIF), and UL 4600 (Safety Case), and others as necessary.”²⁷²

The fifth model of hard law/soft law interaction is for parallel governance by hard law and soft law.²⁷³ An example of such parallel governance recommended in the NHTSA comments was the suggestion of the American Trucking Associations (ATA) that NHTSA and ATA work in partnership to ensure that the ATA best practices align with the NHTSA regulatory requirements, as currently exists in some other trucking safety issues.²⁷⁴ Similarly, the AV developer Motional suggested a program here industry and NHTSA work in parallel.²⁷⁵ Motional began by noting that “[t]here is growing agreement among AV industry stakeholders that process-based measures such as ISO 26262, ISO 21448, IEEE TS17961 and UL 4600 can be used by AV developers to support a safety case.”²⁷⁶ It then recommended that NHTSA could work in parallel with these efforts by “providing holistic guidance for utilizing process measures, including those identified above (ISO 26262, ISO 21448, IEEE TS17961 and UL 4600) and any others that are under development (such as IEEE P2846) or developed in the future, to support a safety case.²⁷⁷ Some of the steps NHTSA could take would include to “establish and maintain a digital repository for available guidance, best practices, and technical standards” and to “develop non-binding guidance on the application and utilization of these process measures by AV developers.”²⁷⁸

Finally, the sixth model is to use hard law to enforce soft law.²⁷⁹ The Insurance Institute for Highway Safety and Highway Loss Data Institute (IIHS-HLDI) proposed a mechanism for such hard law enforcement of soft law conformance.²⁸⁰ The organization recommended that “auditing ADS developers’ safety cases would provide NHTSA with an opportunity to shape ADS safety ahead of its deployment onto our nation’s roads. It seems that ISO2148, ISO26262, UL4600, or some combination of the three could be used as the basis for such an effort.” Aurora suggested a similar approach in its comments in which NHTSA could “identify and organize emerging standards” as well as determine any gaps that exist in the matrix of existing industry standards.²⁸¹

In addition to these six previously described models of hybrid governance, one additional one was suggested in the comments to this NHTSA proceeding. Philip Koopman, on behalf of his consulting firm Edge Case Research, suggested that NHTSA could adopt a hard law rule that requires AV manufacturers to disclose which private AV standards they commit to conform to.²⁸² As discussed above, there is a matrix of private standards promulgated by SAE, IEEE, UL and ISO related to AVs. Many vehicle manufacturer employees participated in the development of these private standards, and many manufacturers refer to such standards as providing relevant guidance. But the companies generally do not publicly disclose if they have committed to conformance with any of these private standards. Koopman’s proposal would require such disclosure,²⁸³ providing a unique model for hard law and soft law to work together.

B. Response to Biden’s Proposed AV STEP Program

Only 37 comments were submitted in response to the Biden AV STEP Proposed rulemaking, of which 31 were submitted by organizations.²⁸⁴ This is far less than the 125 organizational comments (and over 750 total comments) submitted on the Trump ANPRM four years earlier.²⁸⁵ There are several possible reasons for this discrepancy. First, the comment period for the Trump notice extended into the Biden administration, and many of the comments filed were expressly directed to the Biden administration and, in particular, its incoming Secretary of Transportation, Pete Buttigieg. There was a lot of expectation that the Biden administration and Secretary Buttigieg would take regulatory action on AVs, so the comments on the Trump ANPRM were an opportunity for all relevant stakeholders to try to influence any such regulatory initiatives. In contrast, the Biden AV STEP proposal was published just before the second Trump administration took office and, given President Trump’s anti-regulatory philosophy and antipathy to policy initiatives of his predecessors, the AV STEP proposal was very likely dead on arrival, which no doubt reduced interest in submitting comments on the doomed proposal.

In addition, the Trump ANPRM had a much broader and open scope, requesting comments on essentially any aspect of AV governance, including whether NHTSA should focus on hard law or soft approaches. The Biden proposed rule on AV STEP was a more advanced and narrow proposal, requesting comment only on a specific voluntary program described in the proposal. The Biden NPRM dismissed the feasibility of hard law regulations at this time for AVs, so the scope of requested comments was much narrower and may have been of less interest to some stakeholders.

As with the Trump ANPRM, we analyze below the public comments by stakeholder organizations to the Biden AV TEST proposal, including support for hard law versus soft law.

1. Hard Law vs. Soft Law Statistics

The summary statistics for the comments filed in response to the Biden AV STEP NPRM are summarized in Table 2 below. Note that in addition to many fewer overall comments compared to the November 2020 Trump ANPRM, the comments are much more aligned by type of commentator. Almost all the industry commentators (except for one insurance industry trade group) supported a soft law approach, whereas all but one of the NGO comments favored a hard law approach. Overall, we see this strong and approximately even split between hard law and soft law approaches. Relatively fewer organizations expressly supported a hybrid hard law/soft law approach, but the AV STEP program is itself a type of hard law/soft law approach,²⁸⁶and most commentators were supportive of the AV STEP program, although with suggested refinements to make it more or less flexible or rigorous. So, in reality, even though only four commentators proposed new hard law/soft law hybrid approaches, the majority of commentators seemingly (or arguably) were supporting a hybrid approach implicitly.

Table 2: Summary of Comments on Biden AV STAR ANPRM²⁸⁷

Type of Commentator	Number of Total	Favoring Hard Law	Favoring Soft Law	Favoring Hybrid Hard-Soft Law
Industry	11	1 (9%)	10 (91%)	0 (0%)
NGO	6	5 (83%)	1 (17%)	2 (33%)
Professional	5	3 (60%)	1 (20%)	1 (20%)
Governmental	5	2 (40%)	2 (40%)	1 (20%)
Total	27	11 (41%)	14 (52%)	4 (15%)

2. Comments Favoring Hard Law

Almost half (41 percent) of the organizational comments on the Biden AV STEP program favored hard law, even though the proposal itself was primarily based on a voluntary commitment. The Transport Workers Union of America (TWU) was one of the few commentators squarely opposed to the AV STEP proposal, due to its voluntary approach.²⁸⁸ The union stated that “[w]e see nothing in this voluntary program that would entice AV operators to give up their current, unregulated position. The most likely result of this MPRM (sic) would be federal endorsement of unsafe practices and a general degradation of road safety.”²⁸⁹ The TWU was most strongly opposed to the use of independent third-party assessors to evaluate AV safety:

The TWU strongly opposes the scheme proposed in this NPRM to allow so-called “independent assessors” to take responsibility for applicants’ initial safety reviews. This system is rife with pay-to-play incentives and will result in a race to the bottom for these assessors. All NHTSA would accomplish with this proposal would be to enrich the least scrupulous set of academics willing to put their names on tech-company marketing materials thinly veiled as safety studies.²⁹⁰

Similarly, the Transportation Trades Department, AFL-CIO (TTD) commented, “[v]oluntary programs with limited oversight, such as AV STEP, are wholly inadequate to ensure safety and security in transportation.²⁹¹

The National Safety Council (NSC) submitted extensive comments arguing for a mandatory hard law approach rather than the soft law AV STEP proposal.²⁹² The NSC’s primary recommendation is to make AV STEP mandatory rather than voluntary – “[t]o ensure robust public confidence in this program and ADS technology, NSC recommends shifting this program from voluntary to mandatory until the technology fully matures.”²⁹³ Alternatively, if the AV STEP program remains voluntary, the NSC recommends adopting and enforcing a mandatory FMVSS for AVs in tandem with AV STEP.²⁹⁴ NSC also opposes the use of independent third-parties to evaluate manufacturers’ AVs_—rather, this task should be undertaken by NHTSA itself.²⁹⁵ Moreover, the assessment should not be based on private standards such as ANSI/UL 4600, as suggested by NHTSA, because such standards allegedly do not provide a robust assessment of AV safety.²⁹⁶

The National League of Cities supported a hard law approach based on NHTSA developing FMVSSs for AVs, even though NHTSA proclaimed it lacked sufficient evidence for promulgating such standards in its AV SAFE proposal.²⁹⁷ The League did not address NHTSA’s statement about FMVSSs, but recommended that “we should allow the Secretary of Transportation to use NHTSA’s expertise and existing statutory authority to allow for updates to the Federal motor vehicle safety standards (FMVSS) to introduce innovative AVs that are produced and deployed in the U.S. while also protecting the safety of the public through appropriate reporting and limits on the number of vehicles.”²⁹⁸

The Truck Safety Coalition (TSC), Citizens for Reliable and Safe Highways (CRASH), and Parents Against Tired Truckers (PATT) filed a joint comment supporting AV SAFE, but urging that this is just the first step towards issuing “rigorous rules, regulations, guidance, reporting and testing standards necessary to ensure any potential safe deployment of driverless trucks in interstate commerce.”²⁹⁹ These commentators argued that DOT precedent does “not support an uninformed, industry-self-regulated approach.”³⁰⁰ Furthermore, “[a]bsent any federal safety framework, AV trucking companies are essentially self-regulated leaving the public with little recourse but to trust they act in good faith, free from commercial self-interest interference with safety operations.”³⁰¹

3. Comments Favoring Soft Law

Over half (52 percent) of the comments on the AV STEP proposal endorsed a soft law approach. The Alliance for Automotive Innovators, a trade association representing most vehicle manufacturers, submitted extensive comments on the AV STEP proposal.³⁰² The Alliance generally supported the voluntary approach of AV STEP, but proposed a number of refinements to make the program less burdensome on vehicle manufacturers. For example, the Alliance called on NHTSA to use AV STEP to expressly preempt any state requirements that conflict or overlap with the safety design and operation aspects of AV STEP.³⁰³ The Alliance also recommended that manufacturers should be able to self-certify their conformance with private standards rather than having an independent third party perform that function.³⁰⁴ Several Alliance members also filed their own separate comments supporting a soft law approach.³⁰⁵

The National Automobile Dealers Association (NADA) supported a soft law voluntary approach to AV governance. NADA commented: “As NHTSA correctly acknowledges, ‘[v]oluntary programs have historically played an important role in advancing automotive safety, particularly for advanced vehicle technologies.’”³⁰⁶ NADA then added “[a]ccordingly, NADA supports the voluntary approach of AV STEP.”³⁰⁷

The ACES Mobility Coalition also supported a soft law approach, but argued that the AV STAR proposal was too bureaucratic and did not provide adequate incentives for companies to voluntarily op-in.³⁰⁸ The Coalition argues that “the proposed voluntary framework instead introduces unnecessary reporting burdens and new operational restrictions, without providing any meaningful regulatory or safety benefits.”³⁰⁹ Moreover, “AV STEP offers no incentive for participation.”³¹⁰

4. Comments Supporting Hard Law/Soft Law Hybrid Approaches

Only 4 comments (15 percent) specifically endorsed a hard law/soft law hybrid approach. However, this relatively low number must be considered in the context of the AV STEP program, the focus of the 2025 Biden NPRM, which itself is at its core a hybrid hard law/soft law approach. In fact, it presents yet another hard law/soft hybrid model not previously identified in this Article,³¹¹ in which companies are invited to voluntarily opt in to the AV STEP program (a soft law measure), but once they have opted in, they are subject to hard law requirements that are enforceable by NHTSA. Since most commentators supported AV STEP in at least some of its components, even if they also pushed for more soft law or hard law emphasis, one could argue that most commentators implicitly favored a hard law/soft law hybrid approach.

In addition, a few commentators recommended other hybrid hard law/soft law approaches, which fall into several of the categories of hybrid approaches described above.³¹² The National Association of City Transportation Officials (NACTO) favored a hybrid hard law/soft law approach centered on NHTSA using its convening power to create a stakeholder group to develop a comprehensive farmwork for AVs.³¹³ “NACTO recommends organizing a new stakeholder group to bring together cities, states, the U.S. DOT, and private industry to provide input into AV standards on public streets.”³¹⁴ This convening role is one of the hard law/soft hybrid models discussed above.³¹⁵

Underwriter Laboratories (UL), which among other activities develops industry standards, advocated for the roster of private AV standards to be incorporated into the NHTSA AV STEP program.³¹⁶ This incorporation of soft law into hard law is one of the hybrid models discussed above.³¹⁷ The Truck Safety Coalition, although arguing overall for hard law regulation, did support AV STEP as a first step soft law approach towards hard law regulation.³¹⁸ This phasing of soft law to hard law is one of the hard law/soft law models described above.³¹⁹ The National Safety Council, while also supporting a purely hard law approach as its preference, does suggest as an alternative a “tandem” governance system with the voluntary AV STEP operating in coordination with mandatory AV FMVSSs.³²⁰ This proposal aligns with the parallel governance model of hard law/soft law hybrids.³²¹

C. Summary of Public Comments on the Two Proposed Frameworks

In the waning days of their Administrations, NHTSA in the first Trump Presidency and the Biden Presidency published Federal Register notices proposing comprehensive frameworks for AV governance and requesting public comment.³²² The proposals themselves were likely dead letters because of the pending change in administration that had been determined before the proposals were published, but the public comments submitted on each proposal provide a useful snapshot of where stakeholders stand on the difficult issue of AV governance.

The public comments submitted on the two proposals demonstrate that stakeholder organizations are sharply divided on whether AVs should be governed by hard law regulation or soft law measures at this time. Roughly half the commentators ardently supported a hard law approach, arguing that public safety can only be guaranteed by hard rules enforced by NHTSA and that reliance on soft law measures was unacceptable. The other approximately half of organizational stakeholders equally adamantly supported a soft law approach, arguing that there was insufficient evidence to impose hard law rules on this early-stage technology, and that rigorous hard law rules would squash innovation. Given this stalemate, it is unlikely that a consensus will develop for either a purely hard law or soft law approach in the near future.

One ray of hope for a possible greater agreement, however, is combining hard law and soft law in a hybrid approach. Several commentators specifically advocated a hybrid approach, and many other stakeholders are likely to support an intermediate approach. The first part of this Article identified six different hard law/soft law hybrid models from the literature,³²³ and each of these six hybrid models was supported by one or more commentators on the two AV framework proposals.³²⁴ In addition, two additional hybrid models were identified when reviewing those proposals and comments, one from a comment on the Trump ANPRM,³²⁵ and another from the Biden NPRM itself.³²⁶

Table 3: Hybrid Hard Law/Soft Law Models for AV Governance

Model	Description
A	Regulatory agency convenes soft law program
B	Hard law *backstops* soft law program
C	Soft law *evolves* into hard law
D	Hard law *incorporates* soft law
E	Hard law and soft law govern in *parallel*
F	Hard law used to *enforce* soft law commitments
G	Hard law requires *disclosure* of soft law conformance
H	Soft law program *invites* hard law compliance

The eight hybrid models in Table 3 are not exclusive—it is possible for a governance framework to integrate two or more of these models. It is likely that the eventual U.S. governance framework for AVs will involve one or more of these hybrid models.

V. Conclusion

Today, AVs are steadily spreading across the nation, although limited mostly to ride-hailing services in large municipalities. But while the technology proliferates, regulation is limited, confused, and stultified. While everyone from federal agencies to Congress, state agencies and policymakers, consumer groups and think tanks, scholars and the industry itself call for a coordinated regulatory framework, none exists, and frankly, it is hard to see how such a framework based in traditional hard law regulation will emerge.³²⁷

NHTSA, the federal agency with jurisdiction to regulate AV safety, lacks the expertise, resources, and authority to effectively regulate AVs on its own. Its traditional focus and expertise have been on conventional vehicles driven by humans, and it lacks the front-line knowledge and expertise about the diverse technologies and strategies that manufacturers are pursuing in the race for the future AV market.³²⁸ The rapid pace of development and change in AV technologies and business models further challenge NHTSA’s capacity to effectively regulate.³²⁹ This situation has been worsened by the announcement in July 2025 that 25 percent of the already over-extended NHTSA staff have been eliminated due to the current Administration’s cutbacks.³³⁰ Moreover, the primary regulatory authority that NHTSA has for regulating vehicle safety, the FMVSS, is particularly misaligned for regulating AVs. FMVSS must be based on known and objective test criteria, which are easy for AVs to be programmed to master,³³¹ when the real safety concern with AVs is unknown and subjective “corner” or “edge” cases, which by definition cannot be included in FMVSS testing regimes. Congress remains locked in gridlock on this and so many other technology regulatory issues, and regardless it is difficult to conceive what type of legislation they could adopt that would solve the problems of AV regulation. States do not have the authority to regulate the safety design of AVs, and most stakeholders believe that state-by-state regulation will only create a confusing and counter-productive piecemeal mess.

A comprehensive hard law regulatory framework for Avs, therefore, does not seem likely or feasible anytime soon. Alternatively, a voluntary approach emphasizing soft law, which is essentially the status quo, could be continued and strengthened.³³² Soft law has many advantages in terms of being more flexible, agile, cooperative, diverse, and expert-based.³³³ But soft law has its shortcomings as well, including the risk that some manufacturers will not accede to the soft law measures or will not implement them earnestly.³³⁴ Relatedly, the public often does not trust soft law measures, and a lack of consumer trust is already one of the biggest problems facing the AV industry.³³⁵ So it seems a purely soft law approach also may not be acceptable or feasible, and many of the commentators on the two Federal Register notices expressed great frustration with NHTSA’s continued reliance on voluntary approaches.

That leaves some type of hybrid hard law/soft law approach as the only viable approach. Hard law and soft law are often addressed as competing all-or-none approaches to technology governance. However, as this article has explored both in theory and in the context of AV governance, a combination of hard law and soft law is often a sensible approach.³³⁶ Specifically, this Article has identified eight models of hybrid hard law/soft law models. Each of these eight hybrid models has been endorsed by at least one AV stakeholder organization. A hybrid approach may provide a pathway towards AV governance that is acceptable to most stakeholders. Accordingly, it may be productive to focus the AV governance debate on one or some combination of the eight hybrid hard law/soft law models identified in this Article.

* Gary Marchant is Regents Professor and Faculty Director of the Center for Law, Science & Innovation at the Sandra Day O’Connor College of Law at Arizona State University (ASU). Chase Martins is a third-year J.D. candidate at the Sandra Day O’Connor College of Law at ASU. Vrinda Gupta is a 2025 J.D. graduate of the Sandra Day O’Connor College of Law at ASU.