Agentic AI Regulation and Autonomous Vehicles: The Compliance Map Through 2028

Agentic AI Regulation and Autonomous Vehicles: The Compliance Map Through 2028

Executive Summary

Agentic AI regulation reaches autonomous vehicles through vehicle type approval and product safety law, a route that Regulation (EU) 2026/1744 delayed to August 2, 2028 for automotive AI. The general artificial intelligence statutes now moving through American legislatures were drafted for models and for consequential decisions about people, and almost none of them govern the driving task. The binding requirements for a vehicle that plans and executes its own maneuvers come from a narrower set of instruments: the Federal Motor Vehicle Safety Standards and the exemption sections of 49 U.S.C. Chapter 301, the state permitting codes in California and Texas, the United Nations vehicle regulations that Europe and the United Kingdom incorporate by reference, and the type approval rules that sit beneath the European Union artificial intelligence rules.

The picture shifted twice in 2026. On June 26, 2026, the National Highway Traffic Safety Administration withdrew the voluntary oversight program it had proposed eighteen months earlier and committed instead to exemptions and to rewriting the safety standards one at a time. On July 27, 2026, the European Union pushed its high risk obligations for artificial intelligence embedded in regulated products back a full year, to August 2, 2028, and gave the Commission power to displace those obligations entirely where sectoral law already carries equivalent artificial intelligence requirements.

For a manufacturer, the compliance work that matters between now and 2028 sits in vehicle safety files, exemption petitions, permit conditions, and software update management systems. The artificial intelligence statutes operate as liability and disclosure overlays on top of that work. California Civil Code Section 1714.46, effective January 1, 2026, removes the defense that the artificial intelligence autonomously caused the harm. That single sentence will shape autonomous vehicle litigation more than any transparency statute now on the books.

Detailed Findings

What does agentic AI mean inside a vehicle?

Agentic artificial intelligence in a vehicle means a model that perceives, plans, and actuates without a human approving each step, which is what an automated driving system has always done. The term migrated from enterprise software, where an agent books travel or files a support ticket, and it entered vehicle work because the driving stacks shipping in 2026 look less like rule based planners and more like large models trained end to end from sensor input to steering and throttle commands.

The distinction matters for evidence. A rule based planner can be interrogated, because engineers can point to the branch that fired and to the threshold that produced it. A model trained end to end produces a trajectory without an intermediate artifact that a plaintiff’s expert, a regulator, or in house counsel can read. NHTSA engineering analysis EA26002, opened March 18, 2026 and covering approximately 3.2 million Tesla vehicles, turns on that problem directly. The alleged defect is the failure of the system’s own degradation detection to recognize that glare or airborne obscurants had degraded its perception, and to warn the driver in time. Nine incidents, two of them producing injuries and one a fatality, sit behind that file. The investigation upgraded preliminary evaluation PE24031.

A second agentic layer is arriving in cars that has nothing to do with driving. Conversational assistants built on general purpose models now book charging stops, authorize payments, and change vehicle settings. Those functions sit outside the vehicle safety standards and inside consumer protection, payments, and privacy law, and they are the part of the vehicle most likely to be caught by the state artificial intelligence disclosure statutes.

How does federal law regulate an automated driving system today?

Federal law regulates the vehicle, not the driver, and no Federal Motor Vehicle Safety Standard yet tells an automated driving system how to drive. NHTSA sets equipment and performance standards under 49 U.S.C. Chapter 301. States license drivers, register vehicles, write traffic rules, and regulate insurance. An automated driving system falls into the seam between the two. Nothing in the standards prohibits it, and nothing certifies it.

Manufacturers therefore work through exemptions. Section 30113 permits exemption from a standard for a limited number of vehicles. Section 30114(a) permits exemption for vehicles used for research, demonstration, or similar purposes. In April 2025, NHTSA opened its Automated Vehicle Exemption Program to domestically built vehicles under Section 30114(a), an authority previously reserved in practice for imports. On July 31, 2026, the agency published interim guidance on general exemptions in Docket NHTSA-2026-1552 with comments due August 31, 2026, and it replaced static exemption terms lasting a vehicle’s service life with operational authorizations the agency can modify as the system changes.

Crash reporting is the other live federal obligation. The third amended Standing General Order 2021-01 took effect in June 2025 and narrowed reporting to crashes the agency will actually review. From June through December 2025, NHTSA received 517 crash reports involving automated driving systems and 787 involving Level 2 driver assistance features.

What did NHTSA build in 2026, and what did it abandon?

NHTSA abandoned voluntary oversight and committed to rewriting the standards one at a time. On June 26, 2026, the agency withdrew the ADS Equipped Vehicle Safety, Transparency, and Evaluation Program at 91 Fed. Reg. 38619, eighteen months after proposing it at 90 Fed. Reg. 4130 on January 15, 2025. The withdrawal notice records why: industry commenters treated the reporting burden as heavier than the benefit, safety commenters treated a voluntary structure as inadequate for unproven technology on public roads, and the agency concluded the program would not accomplish its objectives.

What replaced it is narrower and harder to argue with. In March 2026, NHTSA proposed modernizing FMVSS No. 102, No. 103, and No. 104 to accommodate vehicles built without manual controls. On June 26, 2026, it proposed the same for FMVSS No. 135 in Docket NHTSA-2026-0728, removing the requirement for a foot operated service brake pedal and a hand or foot operated parking brake, permitting brake activation by electronic command signal, and requiring brake telltales visible to all occupants instead of only the driver position. Stopping distance performance requirements stay. A vehicle with no steering wheel becomes certifiable one standard at a time, and each standard rewritten is a barrier that no longer requires an exemption.

Will Congress preempt state autonomous vehicle rules?

Congress has moved further on autonomous vehicle legislation in 2026 than in any year since 2017, and preemption sits at the center of both live vehicles. The SELF DRIVE Act of 2026, taken up by the House Energy and Commerce subcommittee on January 13, 2026, would require manufacturers of SAE Level 3, Level 4, and Level 5 systems to produce a safety case addressing hardware and software components, operational boundaries, crash anticipation, obstacle detection, and cybersecurity policy, and to submit it to regulators on request. It would bar states and localities from restricting the manufacture, sale, offer for sale, or introduction into interstate commerce of a vehicle whose documentation complies, while leaving room for state rules governing vehicles already deployed. The exemption ceiling of 2,500 vehicles per manufacturer per year would rise, and one proposal circulated in committee sets the figure at 90,000.

The heavy vehicle side moved first. H.R. 8870, the BUILD America 250 Act, cleared the House Committee on Transportation and Infrastructure on May 22, 2026 by a vote of 62 to 2. It directs the Department of Transportation to set performance based safety requirements within two years, under which an automated commercial motor vehicle must perform at least as safely as a human driven vehicle. It confirms that existing federal preemption of state commercial motor vehicle safety law extends to automated commercial vehicles. It also declines to let the Department exempt automated trucks from inspection and surveillance requirements, or human drivers from commercial driver standards, and it imposes hours of service rules on remote and fallback drivers.

How does Executive Order 14365 change the state artificial intelligence picture?

Executive Order 14365, Ensuring a National Policy Framework for Artificial Intelligence, signed December 11, 2025, created a federal mechanism for attacking state artificial intelligence statutes without preempting any of them by its own force. It directed the Attorney General to establish an AI Litigation Task Force within 30 days to challenge state laws on interstate commerce and First Amendment grounds, the Secretary of Commerce to identify within 90 days which state laws the task force should pursue, the Federal Communications Commission to open a proceeding on federal reporting and disclosure standards that would displace conflicting state requirements, and the Federal Trade Commission to issue a policy statement on application of the FTC Act to artificial intelligence models. Eligibility for federal assistance, including broadband equity funds, was tied to whether a state’s artificial intelligence rules align with federal policy. Child safety, artificial intelligence infrastructure, and state procurement were carved out.

For autonomous vehicles the order matters less than the headlines suggest, because state autonomous vehicle codes are vehicle operation rules and not artificial intelligence rules. California permit conditions and Texas authorization requirements regulate who may operate a driverless vehicle on a public road, territory the federal vehicle safety statute has never occupied and that the order does not address. The pressure lands instead on the general artificial intelligence statutes, including the Colorado law the order singled out, and on the disclosure statutes that would reach an in vehicle assistant.

Which state laws now bind an automated driving system?

Two categories bind, and they come from different committees with different theories of the problem.

The first is the autonomous vehicle permitting code. California finalized amended regulations on April 28, 2026 under Title 13 of the California Code of Regulations, ending the prohibition on heavy duty autonomous testing and deployment and imposing a phased path: testing with a safety driver, then driverless testing, then commercial deployment. A heavy duty vehicle must accumulate at least 1,000,000 miles of testing, including 200,000 miles in California, before commercial operation. Manufacturers must report system failures affecting autonomous performance, vehicle immobilizations, collisions, vehicle miles traveled, and traffic law noncompliance. Beginning July 1, 2026, a law enforcement officer may issue a notice of autonomous vehicle noncompliance, and the operator must respond within 72 hours. The Department of Motor Vehicles may suspend or revoke a permit on a finding that operation poses an unreasonable risk of accident, death, or injury, or constitutes an imminent hazard. Fare carrying robotaxi service still requires separate California Public Utilities Commission authority. Texas has required a Department of Motor Vehicles permit since September 1, 2025 under Senate Bill 2807 of the 89th Legislature, implemented through Title 43 of the Texas Administrative Code.

The second category is the general artificial intelligence statute. California Senate Bill 53, the Transparency in Frontier Artificial Intelligence Act, Chapter 138 of the Statutes of 2025, took effect January 1, 2026. It reaches developers that train a foundation model using more than 10 to the 26th cumulative floating point operations, counting fine tuning and material modification, and it imposes heavier duties on developers whose entity and affiliates exceed $500,000,000 in annual gross revenue, including catastrophic risk assessment reporting to the Office of Emergency Services, critical safety incident reporting within fifteen days, notification within twenty four hours where imminent risk of death or serious injury exists, and civil penalties up to $1,000,000 per violation recoverable only by the Attorney General. A driving stack trained at that compute level falls inside the definition even though the statute was written with other harms in view. Colorado replaced its 2024 algorithmic discrimination statute with Senate Bill 189, signed May 14, 2026 and effective January 1, 2027, which covers automated decision technology used in consequential decisions across seven domains and does not reach vehicle operation.

What does the EU artificial intelligence law actually require of automated driving?

Less, and later, than most compliance calendars assume. Regulation (EU) 2024/1689 classifies an artificial intelligence system as high risk when it is a safety component of a product covered by the harmonization legislation listed in Annex I, and Section B of that annex includes the vehicle type approval instruments, Regulation (EU) 2018/858 and Regulation (EU) 2019/2144. Automotive artificial intelligence therefore enters through Section B, where the requirements are taken into account within the existing type approval conformity assessment instead of triggering a separate parallel assessment.

The dates then moved. Regulation (EU) 2026/1744, the Digital Omnibus on AI, entered into force on July 27, 2026 and pushed the Annex I high risk obligations from August 2, 2027 to August 2, 2028. The Annex III use based obligations moved from August 2, 2026 to December 2, 2027, and national regulatory sandboxes slipped to August 2, 2027. The same instrument permits the Commission to limit application of the artificial intelligence duties by delegated act where sectoral law already carries equivalent requirements, and it removed machinery from the full high risk category as the first use of that approach. Whether vehicles follow machinery out of the parallel duties is the most valuable item on an automotive compliance calendar this year.

Type approval for the driving function itself already exists on its own footing. Commission Implementing Regulation (EU) 2022/1426 sets uniform procedures and technical specifications for type approval of the automated driving system of a fully automated vehicle, and it operates whether or not the artificial intelligence duties apply. Directive (EU) 2024/2853 on liability for defective products, with a transposition deadline of December 9, 2026, sits alongside both.

How do the United Nations vehicle regulations set the technical baseline?

They set it for every market that follows the 1958 Agreement, which includes the European Union, the United Kingdom, Japan, and Korea, and they carry the engineering requirements that the artificial intelligence statutes do not. UN Regulation No. 157 governs automated lane keeping systems. UN Regulation No. 155 requires a certified cybersecurity management system across the vehicle lifecycle. UN Regulation No. 156 requires a software update management system. UN Regulation No. 171 covers driver control assistance systems, and the 02 series of amendments to it moved through the Working Party on Automated, Autonomous and Connected Vehicles during 2025 and 2026.

Regulation No. 156 deserves the attention of anyone shipping a learned driving model. A model retrained and pushed to a fleet over the air is a software update to a type approved system, and the update management obligations attach whether the change is a bug fix or an entirely new set of weights. The regulation was written before end to end driving models were commercial, and it now functions as the closest thing the world has to a continuous certification requirement for a changing model.

What are the United Kingdom, Germany, and China doing?

The United Kingdom is building the most complete purpose written statute in force anywhere. The Automated Vehicles Act 2024 creates two licensed roles: the authorised self driving entity, which holds safety responsibility for an authorised vehicle across its life, and the no user in charge operator, which runs vehicles with nobody inside responsible for the driving task. The Automated Vehicles (Permits for Automated Passenger Services) Regulations 2026, SI 2026/439, came into force in the first quarter of 2026 to let commercial pilots begin, carving automated passenger services out of taxi and bus licensing. First passenger pilots were targeted for the second quarter of 2026, consultation on the Statement of Safety Principles for the second and third quarters, draft licensing regulations for the second half of 2026, and full implementation of the Act for the second half of 2027.

Germany went first in Europe on Level 4 and has now gone first on remote operation. Sections 1a through 1l of the Straßenverkehrsgesetz and the AFGBV authorise driverless operation inside an approved operating area. The Straßenverkehr Fernlenk Verordnung, effective December 1, 2025, gives teleoperated driving a uniform legal basis on public roads and treats a remote operator as a driver under Section 1b.

China took the mandatory route. The State Administration for Market Regulation and the National Standardization Administration approved GB 44721 (2026), Safety Requirements for Autonomous Driving Systems of Intelligent Connected Vehicles, on July 30, 2026, effective July 1, 2027. The standard applies to Level 3 and Level 4 systems on passenger and goods vehicles, excluding automated parking. It requires the system to reach safety parity with an attentive human driver, requires Level 3 systems to monitor driver readiness and to manage safe activation and deactivation, and requires manufacturers to maintain safety policy, risk management, safety assurance, and continuous improvement processes, verified through simulation, closed course, and public road testing with third party confirmatory testing. It also requires that automation level, capability, and limitation be communicated in the owner manual, on the website, and on the in vehicle display.

Who pays when an agentic driving system causes harm?

The manufacturer and the operator, and the law has been closing the escape routes.

California Civil Code Section 1714.46, added by Assembly Bill 316 and effective January 1, 2026, provides that a defendant who developed, modified, or used an artificial intelligence system may not assert as a defense that the artificial intelligence autonomously caused the harm. Causation, foreseeability, and comparative fault survive untouched. The autonomy argument does not. The provision reaches the whole supply chain, from foundation model developer through integrator to deploying enterprise, and it does not create strict liability.

In the European Union, Directive (EU) 2024/2853 replaces the 1985 product liability directive and brings software, including artificial intelligence systems and post sale software updates, within the definition of a product. Member State transposition is due December 9, 2026. It adds disclosure obligations and rebuttable presumptions of defect and causation where a claimant faces excessive difficulty proving the case because of technical complexity. For a learned driving model whose internal reasoning cannot be reconstructed after the fact, that presumption decides most cases before the expert testimony begins.

The recall channel is already operating on behavior instead of hardware. Waymo filed a voluntary software recall covering 3,067 vehicles in December 2025, recall 25E084000, addressing fifth generation software that could drive past a stopped school bus displaying flashing red lights and a deployed stop arm. NHTSA had opened its inquiry in October 2025 after an incident in Atlanta. Nothing broke. The vehicle did exactly what its model told it to do, and the model was wrong about a rule of the road that every sixteen year old learns in a classroom.

I spent my early career in a nuclear propulsion plant where the reactor protection system had written setpoints, and any watchstander on the crew could recite the exact condition that would trip the plant. Thirty years later, seated across a negotiating table over supply and indemnity terms for automated driving components, the same question, what condition causes your system to hand control back, often draws a probability distribution instead of a setpoint. Contracts can allocate that uncertainty. They cannot remove it. The indemnity language that survives a real claim is the language that defines the trigger in terms of observable vehicle behavior, logged and timestamped, instead of model confidence.

Want more industry insights? I publish daily analysis on the intersection of deep tech and venture law.

Subscribe to The Innovation Attorney on Substack

Open Questions

Will the Commission use its new delegated act power to pull vehicles out of the parallel artificial intelligence rules the way it pulled machinery? If it does, automotive compliance collapses back into type approval and the August 2, 2028 date loses most of its force. If it does not, manufacturers face two conformity assessments for the same driving function. Investors pricing European automated driving programs should treat this as the largest single regulatory variable in the model.

Does a retrained driving model constitute a new type for type approval purposes, or a software update under UN Regulation No. 156? The answer determines whether a fleet wide model refresh takes days or months, and it therefore determines how fast a European or United Kingdom deployment can improve after launch relative to an American one.

Does California Senate Bill 53 reach an automated driving model trained above 10 to the 26th floating point operations? The statute defines a foundation model by breadth of training data and adaptability across tasks. A driving model trained on general video and adapted to driving may qualify. No enforcement action has tested the question, and a first mover in that enforcement queue will set the compute accounting practice for the industry.

What survives of state autonomous vehicle permit conditions if the SELF DRIVE Act passes with its preemption clause intact? The clause reaches manufacture, sale, and introduction into interstate commerce, and preserves state authority over deployed vehicles. California reporting duties and the notice of autonomous vehicle noncompliance procedure sit on the deployment side of that line, which means the patchwork survives the bill that was written to end it.

Can a plaintiff obtain the training data, evaluation results, and disengagement logs behind a driving model, and under what protective terms? Directive (EU) 2024/2853 answers yes in Europe through its disclosure mechanism. American discovery practice has not yet produced a published ruling on the trade secret boundary for driving model weights and evaluation sets.

Who is the operator when an in vehicle assistant, and not the driving stack, takes an action with legal consequence, such as authorizing a payment or altering a safety setting at a passenger’s request? Neither the vehicle safety standards nor the state autonomous vehicle codes address the question, and the agency law analysis that governs commercial artificial intelligence agents was built for contracting parties, not for moving vehicles.

Source List

Primary legal sources. Regulation (EU) 2024/1689 of 13 June 2024 laying down harmonised rules on artificial intelligence, OJ L, 12 July 2024. Establishes the high risk classification route through Annex I Section B that captures automotive artificial intelligence.

Regulation (EU) 2026/1744, the Digital Omnibus on AI, in force 27 July 2026. Moves the Annex I high risk compliance date to 2 August 2028 and authorises the Commission to displace artificial intelligence duties where sectoral law is equivalent.

Commission Implementing Regulation (EU) 2022/1426 of 5 August 2022 on type approval of the automated driving system of fully automated vehicles. Sets the operative European technical procedure for approving a driving function.

Directive (EU) 2024/2853 of 23 October 2024 on liability for defective products, transposition due 9 December 2026. Brings software and artificial intelligence within the product definition and adds presumptions of defect and causation.

California Civil Code Section 1714.46, added by Assembly Bill 316 (2025), effective 1 January 2026. Bars the defense that artificial intelligence autonomously caused the harm.

California Senate Bill 53, Transparency in Frontier Artificial Intelligence Act, Chapter 138, Statutes of 2025, effective 1 January 2026. Imposes transparency, catastrophic risk reporting, and incident reporting duties on frontier developers.

Colorado Senate Bill 189 (2026), signed 14 May 2026, effective 1 January 2027. Replaces the 2024 algorithmic discrimination statute with a transparency and disclosure statute for automated decision technology.

Texas Senate Bill 2807 (89th Legislature, 2025), effective 1 September 2025, and Title 43, Texas Administrative Code. Requires a state permit for driverless operation.

Title 13, California Code of Regulations, autonomous vehicle testing and deployment regulations finalized 28 April 2026. Authorises heavy duty autonomous operation and adds enforcement and reporting duties.

Automated Vehicles Act 2024 (United Kingdom) and the Automated Vehicles (Permits for Automated Passenger Services) Regulations 2026, SI 2026/439. Create the authorised self driving entity and no user in charge operator roles and the passenger service permit.

GB 44721 (2026), Safety Requirements for Autonomous Driving Systems of Intelligent Connected Vehicles, approved 30 July 2026, effective 1 July 2027. China’s first mandatory national standard for Level 3 and Level 4 systems.

UN Regulations No. 155, No. 156, No. 157, and No. 171, United Nations Economic Commission for Europe. Govern cybersecurity management, software update management, automated lane keeping, and driver control assistance.

Agency actions and rulemakings. ADS Equipped Vehicle Safety, Transparency, and Evaluation Program; Withdrawal, 91 Fed. Reg. 38619 (26 June 2026), Docket NHTSA-2024-0100. Records the end of the voluntary oversight program and the reasons for it.

AV Framework Updates and Request for Comments on Interim Guidance, 31 July 2026, Docket NHTSA-2026-1552. Sets out the Part 555 interim guidance and the operational authorization structure.

Federal Motor Vehicle Safety Standards; Modernization of FMVSS No. 135 To Accommodate ADS Equipped Vehicles, 26 June 2026, Docket NHTSA-2026-0728. Proposes removing manual brake control requirements while retaining stopping distance performance.

NHTSA Engineering Analysis EA26002, opened 18 March 2026, and recall 25E084000, December 2025. The first covers degradation detection failure across approximately 3.2 million vehicles; the second covers a behavioral software defect in 3,067 robotaxis.

Executive Order 14365, Ensuring a National Policy Framework for Artificial Intelligence, 11 December 2025. Directs federal litigation, funding, and rulemaking pressure against state artificial intelligence statutes.

Legislation in progress. H.R. 8870, the BUILD America 250 Act, reported by the House Committee on Transportation and Infrastructure 22 May 2026, and the SELF DRIVE Act of 2026, heard in the House Energy and Commerce subcommittee 13 January 2026. Both carry preemption provisions bearing on state autonomous vehicle authority.

Secondary sources. Contemporary law firm analyses of the NHTSA automated vehicle framework, the California autonomous vehicle rulemaking, the Digital Omnibus amendments, United Kingdom implementation timing, and United States accountability doctrine for artificial intelligence agents are collected in the bibliography below, with publication dates, and were used to cross check the primary instruments cited above.

Bibliography

Automated Vehicles Act 2024, c. 10 (UK).

Automated Vehicles (Permits for Automated Passenger Services) Regulations 2026, SI 2026/439 (UK). https://www.legislation.gov.uk/uksi/2026/439/contents/made

Cal. Civ. Code Section 1714.46 (added by A.B. 316, 2025 Cal. Stat.).

Cal. Code Regs. tit. 13, Sections 227 to 228 (2026).

Colorado Senate Bill 26-189 (2026).

Commission Implementing Regulation (EU) 2022/1426 of 5 August 2022, OJ L 221/1.

Covington & Burling LLP, California’s New AV Rules Open Door to Heavy Duty Deployment While Imposing Significant New Compliance Obligations (May 29, 2026). https://www.insideglobaltech.com/2026/05/29/californias-new-av-rules-open-door-to-heavy-duty-deployment-while-imposing-significant-new-compliance-obligations/

Crowell & Moring LLP, NHTSA Proposes AV Brake Standard Updates and Withdraws AV STEP Program (2026). https://www.crowell.com/en/insights/client-alerts/nhtsa-proposes-updates-to-federal-brake-standards-for-autonomous-vehicles-and-withdraws-av-step-program

Directive (EU) 2024/2853 of the European Parliament and of the Council of 23 October 2024 on liability for defective products, OJ L, 18 November 2024.

Exec. Order No. 14365, Ensuring a National Policy Framework for Artificial Intelligence, 90 Fed. Reg. (Dec. 11, 2025).

GB 44721 (2026), Safety Requirements for Autonomous Driving Systems of Intelligent Connected Vehicles (State Administration for Market Regulation and National Standardization Administration, approved July 30, 2026).

National Highway Traffic Safety Administration, ADS Equipped Vehicle Safety, Transparency, and Evaluation Program; Withdrawal, 91 Fed. Reg. 38619 (June 26, 2026). https://www.federalregister.gov/documents/2026/06/26/2026-12980/ads-equipped-vehicle-safety-transparency-and-evaluation-program-withdrawal

National Highway Traffic Safety Administration, AV Framework Updates and Request for Comments on Interim Guidance, Docket No. NHTSA-2026-1552 (July 31, 2026). https://www.federalregister.gov/documents/2026/07/31/2026-15483/av-framework-updates-and-request-for-comments-on-interim-guidance

National Highway Traffic Safety Administration, Federal Motor Vehicle Safety Standards; Modernization of FMVSS No. 135 To Accommodate ADS Equipped Vehicles, Docket No. NHTSA-2026-0728 (June 26, 2026). https://www.federalregister.gov/documents/2026/06/26/2026-12981/federal-motor-vehicle-safety-standards-modernization-of-fmvss-no-135-to-accommodate-ads-equipped

National Highway Traffic Safety Administration, Office of Defects Investigation, Engineering Analysis EA26002 (Mar. 18, 2026). https://static.nhtsa.gov/odi/inv/2026/INOA-EA26002-10023.pdf

Regulation (EU) 2018/858 of 30 May 2018 on the approval and market surveillance of motor vehicles, OJ L 151/1.

Regulation (EU) 2019/2144 of 27 November 2019 on type approval requirements for general safety, OJ L 325/1.

Regulation (EU) 2024/1689 of 13 June 2024 laying down harmonised rules on artificial intelligence, OJ L, 12 July 2024.

Regulation (EU) 2026/1744, Digital Omnibus on AI (in force July 27, 2026).

Tex. S.B. 2807, 89th Leg., R.S. (2025); 43 Tex. Admin. Code ch. 220.

United Nations Economic Commission for Europe, UN Regulations Nos. 155, 156, 157, and 171, and Working Party on Automated, Autonomous and Connected Vehicles proceedings (2025 to 2026). https://unece.org/transport/road-transport/working-party-automatedautonomous-and-connected-vehicles-introduction

Interested in analysis about the intersection of tech, policy and the law? Check out my Substack channel. https://theinnovationattorney.substack.com/