Edge AI device legal requirements and the Tesla $329 million Autopilot verdict of 2025.
Edge AI Devices: Commercial Opportunities and Legal Issues for Moving and Stationary Deployments
Executive Summary
Edge AI devices deployed under the Federal Trade Commission Act, the National Traffic and Motor Vehicle Safety Act, and the Food, Drug, and Cosmetic Act face materially different legal obligations depending on whether they operate in moving platforms such as autonomous vehicles and commercial drones or in stationary installations such as hospital imaging systems and industrial manufacturing lines. The global edge AI market reached $24.91 billion in 2025 and is projected to reach $118.69 billion by 2033, growing at a compound annual rate of 21.7%. That growth trajectory is already testing regulatory structures built for a pre-AI world, and the legal exposure in both deployment categories is substantial and accelerating.
For moving devices, the central legal challenge is product liability in the absence of a human operator. A Miami jury awarded $329 million in August 2025 against Tesla for an Autopilot-related fatality, the first verdict finding a driver-assistance AI system defective under products liability doctrine, and that decision opens a new front for manufacturers, fleet operators, and software developers who embed AI at the edge of moving platforms. Regulatory authority over moving devices is shared between the National Highway Traffic Safety Administration, the Federal Aviation Administration, and a patchwork of state licensing regimes, none of which has yet produced a unified federal preemption standard.
For stationary devices, the central legal challenge is compliance across multiple overlapping regimes: the Food and Drug Administration cleared 1,451 cumulative AI-enabled medical devices through the end of 2025, almost all via the 510(k) pathway, while simultaneously advancing a Quality Management System Regulation aligned with ISO 13485 that became effective February 2, 2026. Industrial edge AI deployments face parallel obligations under the Occupational Safety and Health Act, state privacy statutes including the California Consumer Privacy Act regulations effective January 1, 2026, and the EU AI Act’s high-risk system mandates enforceable August 2, 2026.
The intellectual property layer complicates both categories. Edge AI systems run trained model weights on local hardware, and those weights are increasingly the most commercially valuable asset in the stack. Trade secret protection is the primary available mechanism, but courts have not yet resolved whether model weights derived from training on third-party data can be protected as a trade secret when the underlying data may itself carry IP claims. Export controls add a further dimension: the Commerce Department’s Bureau of Industry and Security revised Export Administration Regulations controls on advanced computing and added new controls on AI model weights for certain advanced items in 2025, and the Defense Technology Security Administration signaled additional United States Munitions List revisions for 2026.
Detailed Findings
Part I: Moving Devices
What commercial opportunities does Edge AI create in moving vehicles and drones?
The commercial case for edge AI in moving platforms rests on latency: a connected vehicle relying on a cloud inference call cannot tolerate the 50 to 200 millisecond round-trip time that cloud processing requires when making a collision avoidance decision at highway speeds. On-device inference eliminates that dependency and enables real-time processing of data from cameras, lidar, radar, and ultrasonic sensors without a persistent network connection. That technical reality drives a discrete set of commercial opportunities that differ meaningfully across vehicle classes.
In passenger vehicles, edge AI powers the advanced driver assistance systems now standard on most new vehicles sold in the United States, and it underpins the Level 3 and Level 4 automated driving systems being commercially deployed by Waymo, Zoox, and others. Waymo completed more than four million fully autonomous rides by early 2025 and continues expanding its commercial robotaxi service in San Francisco, Phoenix, and Los Angeles. The commercial model for autonomous ride-hailing requires edge AI that can handle the full dynamic driving task, defined by SAE International Standard J3016 as all real-time operational and tactical functions required to operate a vehicle, without any expectation of driver intervention.
In commercial trucking, Aurora Innovation achieved driverless commercial freight operations on public Texas highways in April 2024, with a network covering Dallas to Houston. The commercial opportunity in long-haul autonomous trucking is estimated at $40 billion annually in the United States based on driver cost alone, and edge AI is the enabling technology because federal highway infrastructure does not provide the dedicated connectivity that cloud-dependent systems require.
In commercial drone operations, the Federal Aviation Administration’s Part 108 BVLOS (Beyond Visual Line of Sight) rulemaking, expected in final form in 2026, will open the commercial drone delivery market to companies including Amazon Prime Air and Wing Aviation. The edge AI requirement is identical to ground vehicles: real-time object detection and collision avoidance that cannot depend on a cellular network connection with variable latency.
In maritime applications, autonomous surface vessels and underwater vehicles for offshore energy inspection, port logistics, and scientific survey represent a growing commercial sector. The International Maritime Organization’s Maritime Autonomous Surface Ships guidelines are being incorporated into flag state regulations, and the commercial opportunity in offshore wind farm inspection alone is projected at $3 billion globally by 2030.
What are the product liability rules for autonomous vehicles using Edge AI?
The Tesla $329 million jury verdict entered in August 2025 in a Miami federal court is the first products liability finding that an AI-powered driver assistance system was defective, and it restructures the risk calculus for every company embedding edge AI in a moving platform. The jury in that case attributed 33% of fault for a fatal April 2019 collision to Tesla, awarding $129 million in compensatory damages and $200 million in punitive damages, with the remaining 67% attributed to the human driver who diverted his attention from the road.
Traditional motor vehicle products liability analysis under the Restatement (Third) of Torts: Products Liability requires proof of a manufacturing defect, a design defect, or a failure to warn. Edge AI in autonomous vehicles creates doctrinal pressure on all three prongs. A manufacturing defect claim targets the specific hardware unit; a design defect claim targets the AI model architecture and training methodology; and a failure-to-warn claim targets how the system’s capabilities and limitations are communicated to operators. The Tesla verdict rested primarily on design defect and failure-to-warn theories: the plaintiff argued that Autopilot was activated on road types for which it was not designed and that Tesla’s marketing overstated its capabilities.
The ’empty driver’s seat’ problem is the most significant doctrinal gap in current tort law applied to edge AI moving devices. When a vehicle operating at SAE Level 4 automation is involved in a collision, there may be no human driver to whom negligence can be attributed. Courts have not yet resolved whether the vehicle manufacturer, the ADS software developer, the fleet operator, the mapping data provider, or some combination of these parties bears the duty of care that a human driver would otherwise owe. Separate liability is allocated differently in the Robinson v. Uber Technologies, Inc. line of cases addressing TNC driver liability, but those cases involve a human driver in the seat and do not resolve the fully automated scenario.
NHTSA published a Notice of Proposed Rulemaking in the Federal Register on January 15, 2025, 90 FR 4130, proposing the ADS-Equipped Vehicle Safety, Transparency, and Evaluation Program, known as AV STEP. AV STEP would establish a voluntary federal oversight framework requiring ADS manufacturers and fleet operators to submit independent safety case assessments, conform to identified industry standards, and file periodic and event-triggered safety reports with the agency. As of June 2026 the agency has not issued a final rule, leaving the regulatory standard of care undefined at the federal level. That gap matters for products liability: without a federal safety standard against which to measure an ADS’s design, plaintiffs may more readily argue that any available alternative design would have prevented the harm.
State-level regulation adds further complexity. California requires ADS operators to submit disengagement reports, is transitioning to documentation of dynamic driving task system failures, and requires safety cases for commercial operations. Texas requires permits for fully autonomous commercial vehicles. Florida has adopted a permissive framework that preempts local restrictions on ADS operations. The resulting patchwork means that a national ADS fleet must comply with 50 potentially different state standards, none of which preempts federal tort law.
Insurance structures for moving edge AI platforms remain in development. Most states require that a vehicle carry minimum liability insurance, but standard auto policies exclude coverage for vehicles operating in an automated mode. Commercial insurers including Munich Re and Swiss Re have developed ADS-specific policies, but pricing and coverage terms are not yet standardized, and the allocation of subrogation rights between the insurer of the vehicle and the manufacturer of the ADS system has not been resolved by any published decision.
What data privacy rules apply to connected vehicles and drones collecting sensor data?
Moving edge AI platforms are persistent data collection devices. A modern autonomous vehicle processes sensor data at rates exceeding one terabyte per hour and may collect information about the vehicle’s occupants, the road environment, and identifiable third parties such as pedestrians. A commercial delivery drone equipped with a camera collects imagery of private property and identifiable individuals below its flight path.
At the federal level, no single comprehensive privacy statute governs connected vehicle data collection. The Federal Trade Commission Act, 15 U.S.C. Section 45, prohibits unfair or deceptive acts or practices and has been interpreted by the FTC to require that data collection practices disclosed to consumers be honored. The FTC’s 2014 Connected Cars report identified data minimization, transparency, and security as core expectations for vehicle data collection, and the agency’s enforcement authority under Section 5 remains the primary federal backstop.
California’s Consumer Privacy Act, as amended by the California Privacy Rights Act and implementing regulations effective January 1, 2026, applies to connected vehicle operators meeting the statutory thresholds. The 2026 regulations require a privacy risk assessment before initiating processing that presents a significant risk to consumer privacy. Vehicle telematics data collected while a California resident is in or near the vehicle is personal information under the CPRA. California Assembly Bill 1008, effective January 1, 2025, explicitly extended CPRA protections to AI-generated data, meaning that inferences drawn from sensor data about a vehicle occupant’s behavior or location are also personal information subject to the Act’s access, deletion, and opt-out rights.
Biometric data collected by interior driver monitoring cameras presents additional exposure under the Illinois Biometric Information Privacy Act, 740 ILCS 14/1, which requires prior written consent before collecting a retinal scan, fingerprint, facial geometry, or voiceprint. Several class actions have been filed against vehicle manufacturers under BIPA for facial recognition features in in-cabin driver monitoring systems, with damages structured at $1,000 per negligent violation and $5,000 per intentional violation. Given that a vehicle sold with active driver monitoring collects biometric data on every driver, the per-device exposure under BIPA is substantial.
What export control obligations apply to AI systems in drones and defense vehicles?
Moving edge AI platforms with military applications are subject to International Traffic in Arms Regulations, 22 C.F.R. Parts 120 to 130, when they qualify as defense articles or incorporate defense services as defined in the United States Munitions List. Autonomous targeting algorithms and fire control systems integrated into military ground vehicles and unmanned aerial systems are clearly within USML Categories VIII and XIX. The more complex question, which DDTC has not fully resolved as of June 2026, is whether AI model weights trained on defense-related data constitute technical data under ITAR, triggering license requirements even for transfer of software files without any physical hardware.
Dual-use moving platforms, such as commercially available drones modified for surveillance or logistics functions that could serve military purposes, fall under the Export Administration Regulations administered by the Commerce Department’s Bureau of Industry and Security. The Commerce Department’s January 2025 Framework for Artificial Intelligence Diffusion, published in the Federal Register at 90 FR 4608, revised EAR controls on advanced computing integrated circuits and added new controls on AI model weights for advanced AI items. A drone manufacturer exporting a vehicle with on-board edge AI inference hardware must classify the hardware under the appropriate Export Control Classification Number and determine whether a license is required for the destination country.
Part II: Stationary Devices
What commercial opportunities does Edge AI create in hospitals, labs, and industrial facilities?
The commercial case for edge AI in stationary devices rests on data sovereignty and latency. A hospital cannot transmit patient imaging data to a commercial cloud for inference without satisfying the Health Insurance Portability and Accountability Act’s minimum necessary standard and executing a Business Associate Agreement with the cloud operator under 45 C.F.R. Part 164. An on-premises edge AI system that runs inference on the device where the data was generated eliminates the transmission risk entirely and reduces inference latency to milliseconds, which matters for real-time clinical decision support at the point of care.
The FDA cleared 1,451 cumulative AI-enabled medical devices through the end of 2025, with radiology imaging representing 76% of all cleared devices. The commercial opportunity in clinical edge AI extends beyond radiology: cardiovascular imaging, pathology slide analysis, neurology imaging, and hospital workflow automation represent categories where FDA clearance rates grew substantially in 2025. The addition of Predetermined Change Control Plans, known as PCCPs, to approximately 10% of 2025 510(k) clearances creates a new commercial model in which a cleared device can update its AI model within a pre-approved envelope without a new clearance submission, reducing the regulatory cost of continuous model improvement.
In industrial manufacturing, edge AI running on-premises inference enables real-time quality inspection, predictive maintenance, and process optimization without exposing proprietary manufacturing data to third-party cloud providers. A semiconductor fabrication facility processing wafers worth $50,000 each cannot tolerate the latency or connectivity dependency of cloud-based defect detection; edge AI inference on local GPU hardware provides the sub-millisecond decision times required for inline process control. The industrial edge AI segment is projected to grow at a compound annual rate of 25.3% through 2030, driven by the manufacturing sector’s shift toward autonomous production lines.
In research laboratories, edge AI enables accelerated scientific discovery by running large model inference on sensitive research data that cannot leave the laboratory network due to export control, grant agreement, or sponsor confidentiality requirements. The Department of Energy’s national laboratory system, including Argonne, Oak Ridge, and Lawrence Berkeley, has deployed on-premises AI inference infrastructure to support materials science, genomics, and climate modeling research that requires processing petabytes of data without cloud transfer. University research laboratories receiving Department of Defense or National Institutes of Health funding face specific data handling obligations under grant terms that may prohibit processing on commercial cloud infrastructure.
What FDA regulatory requirements apply to Edge AI in hospital and clinical settings?
The FDA regulates software that meets the definition of a medical device under 21 U.S.C. Section 321(h), and edge AI clinical decision support tools that are intended to diagnose, treat, cure, or prevent disease meet that definition. The agency’s 2019 Proposed Regulatory Framework for Modifications to Artificial Intelligence and Machine Learning-Based Software as a Medical Device describes a lifecycle-based approach to AI regulation; as of 2026, the primary clearance pathway remains the 510(k) substantial equivalence process under 21 U.S.C. Section 360(k).
The FDA’s Quality Management System Regulation, 21 C.F.R. Part 820, was amended in a final rule published September 2024 to align with ISO 13485:2016, with an effective date of February 2, 2026. Medical device manufacturers who incorporate edge AI inference into their devices must now demonstrate design controls, risk management processes consistent with ISO 14971, and post-market surveillance systems that monitor AI model performance continuously after clearance. For edge AI systems that update their models over time, the PCCP pathway requires a pre-approved description of the types and boundaries of anticipated modifications, with the agency retaining authority to require a new submission if a proposed modification falls outside the approved envelope.
Software as a Medical Device that is intended to replace the clinical judgment of a licensed clinician, rather than support it, is classified by FDA as high risk and requires a Pre-Market Approval application under 21 U.S.C. Section 360e rather than 510(k) clearance. The distinction between a tool that provides information to a clinician and a tool that makes an autonomous clinical decision is the central regulatory dividing line for clinical edge AI, and manufacturers who blur that line in marketing materials face both FDA enforcement action and false advertising liability under 15 U.S.C. Section 45.
Hospital operators who deploy edge AI devices bear compliance obligations independent of the device manufacturer. The Centers for Medicare and Medicaid Services conditions of participation, 42 C.F.R. Part 482, require hospitals to have a quality assessment and performance improvement program that addresses the performance of clinical decision support tools. A hospital that deploys an edge AI imaging tool without adequate clinical validation and oversight policies faces both CMS conditions of participation deficiencies and potential liability for AI-related adverse patient outcomes.
What HIPAA and state privacy obligations apply to Edge AI in healthcare settings?
The Health Insurance Portability and Accountability Act, 45 C.F.R. Parts 160 and 164, governs the use and disclosure of protected health information by covered entities and their business associates. An edge AI device deployed within a covered entity’s facility that processes protected health information in performing its function is operating under the covered entity’s HIPAA compliance obligations. The device manufacturer, if it has access to de-identified data for training or performance monitoring purposes, will be a business associate and must execute a Business Associate Agreement meeting the requirements of 45 C.F.R. Section 164.504(e).
HIPAA’s Security Rule, 45 C.F.R. Section 164.312, requires covered entities to implement technical safeguards to guard against unauthorized access to protected health information transmitted over electronic communications networks. An edge AI device that transmits model updates, telemetry data, or inference logs to a manufacturer’s server is transmitting electronic protected health information if those transmissions contain or could be used to reconstruct patient-specific data. The Security Rule requires encryption of all such transmissions, access controls on the device itself, and audit logs of all access to protected health information.
State biometric privacy statutes add a layer of obligation not preempted by HIPAA. Illinois BIPA, 740 ILCS 14/1, applies to the collection of facial geometry by hospital-based AI systems used for patient identification, surgical planning, or dermatology analysis. Texas’s Capture or Use of Biometric Identifier statute, Texas Business and Commerce Code Chapter 503, and Washington’s My Health MY Data Act, effective for businesses meeting its thresholds, impose similar collection consent and data handling requirements. A hospital deploying edge AI that processes facial or biometric data must obtain prior written consent from the patient and must establish a retention and destruction schedule for the biometric data consistent with each applicable state statute.
What industrial safety and OSHA obligations apply to Edge AI in manufacturing?
The Occupational Safety and Health Act, 29 U.S.C. Section 654, requires employers to furnish a workplace free from recognized hazards likely to cause death or serious physical harm. An edge AI system that controls a robotic arm, automated guided vehicle, or machine tool in a manufacturing facility is a component of the workplace, and its malfunction or unexpected behavior is a recognized hazard that triggers the general duty clause. OSHA has not yet promulgated a standard specifically addressing AI in the workplace; enforcement therefore proceeds under the general duty clause with reference to ANSI/RIA R15.06 and ISO 10218 as industry consensus standards for industrial robot safety.
Edge AI systems that replace or supplement human decision-making in safety-critical process control, such as chemical plant pressure management or electrical grid switching, are subject to the Process Safety Management standard, 29 C.F.R. Section 1910.119, if they are associated with covered chemicals above threshold quantities. The management of change requirements of Section 1910.119(l) require that any change to a process, including the introduction of an AI control system, be reviewed and authorized before implementation, with documentation of the basis for safety of the change.
What intellectual property protections cover Edge AI model weights and architectures?
Edge AI systems deployed on-premises or on-device present IP protection challenges that differ from cloud-deployed AI. Model weights embedded in device firmware are physically present on hardware that may be in the possession of a customer, a competitor, or a hostile actor. The primary available protection is trade secret law under the Defend Trade Secrets Act, 18 U.S.C. Section 1836, which federalizes trade secret misappropriation claims and provides for injunctive relief, compensatory damages, and exemplary damages up to two times actual damages for willful misappropriation.
To establish trade secret protection for model weights under the DTSA, the owner must demonstrate that the weights derive independent economic value from not being generally known and that reasonable measures were taken to maintain secrecy. For edge devices, the reasonable measures requirement necessitates hardware-level protections: encrypted storage for model weights, trusted execution environments that prevent weight extraction, and end-user license agreement provisions prohibiting reverse engineering. Courts have held in several post-2022 decisions that software subject to EULA restrictions can qualify as a trade secret even when distributed to end users, provided the technical measures against extraction are genuine.
Patent protection for edge AI architectures is available but limited. The Supreme Court’s decision in Alice Corp. v. CLS Bank International, 573 U.S. 208 (2014), and the USPTO’s 2019 Revised Guidance on the Application of Section 101, continue to create uncertainty about patent eligibility for AI-implemented inventions. Patent applications claiming AI methods must demonstrate that the claims are directed to a specific technical improvement rather than an abstract idea; claims directed to the architecture of a neural network as implemented in hardware have fared better than method claims directed to the mathematical operations alone.
What EU AI Act obligations apply to stationary Edge AI in hospitals and industrial facilities?
The EU AI Act, Regulation (EU) 2024/1689, entered into force August 1, 2024, with its high-risk AI system obligations becoming enforceable August 2, 2026. Annex III of the Act lists categories of AI systems that are classified as high risk by default, including AI systems used as safety components of medical devices, AI systems for worker management and access to employment, and AI systems used in the management of critical infrastructure. An edge AI device deployed in an EU-market hospital that provides clinical decision support or automated diagnostic outputs is a high-risk AI system subject to the full compliance regime under Chapter III of the Act.
High-risk AI system obligations under the EU AI Act include: a risk management system maintained throughout the system lifecycle under Article 9; data governance requirements under Article 10; technical documentation maintained under Annex IV; automatic logging under Article 12; transparency and information obligations to deployers under Article 13; human oversight measures under Article 14; and a quality management system under Article 17. Before placing a high-risk AI system on the EU market, the provider must complete a conformity assessment under Article 43 and register the system in the EU database under Article 49. Non-compliance exposes providers to fines of up to 30 million euros or 6% of total worldwide annual turnover, whichever is higher, for violations of the high-risk requirements.
The Act applies to providers who place AI systems on the EU market, meaning a US manufacturer of industrial edge AI who sells its product into the EU through a distributor is a provider subject to the Act. The Act also applies to deployers who operate high-risk AI systems within the EU: a German automotive manufacturer that deploys a US-made edge AI quality inspection system on its production line is a deployer with independent obligations under Articles 26 and 27, including conducting a fundamental rights impact assessment before deployment and monitoring the system for risks and incidents.
Open Questions
1. Federal preemption of state ADS liability regimes: Several federal circuit courts have not yet resolved whether NHTSA’s eventual final AV STEP rule will preempt state tort claims arising from ADS failures, or whether a vehicle manufacturer’s compliance with the federal voluntary program establishes a compliance defense. The answer will determine whether a single national liability standard for autonomous vehicles is possible or whether manufacturers face 50 different standards of care.
2. AI model weights as technical data under ITAR: DDTC has not issued definitive guidance on whether AI model weights trained on defense-related data constitute technical data under ITAR’s definition at 22 C.F.R. Section 120.31. Until that guidance issues, companies transferring model weights across borders for edge device deployment in dual-use moving platforms face meaningful license compliance risk.
3. HIPAA applicability to de-identified training data: The FDA’s PCCP pathway permits edge AI medical devices to update their models using post-market performance data. Whether that performance data, even when de-identified under HIPAA’s Safe Harbor standard at 45 C.F.R. Section 164.514(b), can be used for model training by the manufacturer without a Business Associate Agreement is an unresolved question that the HHS Office for Civil Rights has not addressed in formal guidance.
4. OSHA rulemaking on AI in the workplace: OSHA published an Advance Notice of Proposed Rulemaking in 2024 on the use of AI in the workplace but has not yet proposed a specific standard. Until a standard issues, enforcement under the general duty clause leaves manufacturers of industrial edge AI systems uncertain about the specific technical specifications required to satisfy OSHA’s recognized-hazard standard.
5. EU AI Act extraterritorial reach for edge devices: The EU AI Act’s application to providers who place products on the EU market raises an unresolved question about whether a US company that provides edge AI software updates to EU-deployed devices without selling into the EU directly is subject to the Act’s provider obligations. The European AI Office has not yet issued guidance on this specific scenario.
6. Tort liability allocation between ADS manufacturers and fleet operators: Several states have enacted statutes providing that the manufacturer of an ADS system, rather than the registered owner of the vehicle, bears the liability of a driver when the system is in operation. These statutes have not been tested in appellate courts, and no court has addressed whether indemnification agreements between manufacturers and fleet operators are enforceable as against the public policy expressed in those statutes.
7. Trade secret protection for model weights in open-weight deployments: Several edge AI platforms use open-weight models such as Meta’s Llama series as a base and apply proprietary fine-tuning. Whether the fine-tuned weights constitute a protectable trade secret separate from the base open-weight model, and how to define the secrecy and value requirements when the base is publicly available, has not been addressed in any published decision under the DTSA.
Source List
National Highway Traffic Safety Administration, ADS-Equipped Vehicle Safety, Transparency, and Evaluation Program (AV STEP), Notice of Proposed Rulemaking, 90 FR 4130 (January 15, 2025). Primary regulatory source for federal ADS oversight framework.
Commerce Department, Bureau of Industry and Security, Framework for Artificial Intelligence Diffusion, 90 FR 4608 (January 15, 2025). Primary source for EAR controls on AI model weights.
EU Artificial Intelligence Act, Regulation (EU) 2024/1689, Official Journal of the European Union (August 1, 2024). Primary EU regulatory source for high-risk AI system obligations.
Food and Drug Administration, Quality Management System Regulation, 21 C.F.R. Part 820, Final Rule, 89 FR 70396 (September 2024). Primary source for medical device quality requirements effective February 2, 2026.
Alice Corp. v. CLS Bank International, 573 U.S. 208 (2014). Controlling Supreme Court decision on patent eligibility for AI-implemented inventions.
Defend Trade Secrets Act, 18 U.S.C. Section 1836 (2016). Primary federal trade secret statute applicable to model weight protection.
Illinois Biometric Information Privacy Act, 740 ILCS 14/1. State statute governing biometric data collection by edge AI systems with cameras.
Anzalone and Doyle Trial Lawyers, ‘Jury Finds Tesla Autopilot Defective in Landmark $329 Million Verdict,’ August 2025. Source for first jury verdict finding ADS defective under products liability doctrine.
Eno Center for Transportation, ‘2025 Autonomous Vehicles Federal Policy Wrapped,’ December 2025. Analysis of federal AV regulatory developments in 2025.
Grand View Research, ‘Edge AI Market Size, Share and Trends Analysis Report,’ 2025. Market sizing data for global edge AI sector.
Covington and Burling LLP, ‘NHTSA Proposes New Autonomous Vehicle Program,’ January 2025. Legal analysis of AV STEP NPRM.
Innolitics, ‘2025 Year in Review: AI and ML Medical Device 510(k) Clearances,’ December 2025. Statistical source for FDA AI device clearance data.
Bibliography
Alice Corp. v. CLS Bank International, 573 U.S. 208 (2014).
Anzalone and Doyle Trial Lawyers. ‘Jury Finds Tesla Autopilot Defective in Landmark $329 Million Verdict: What It Means for Future Cases.’ August 2025. https://www.anzalonelaw.com/jury-finds-tesla-autopilot-defective-in-landmark-329-million-verdict-what-it-means-for-future-cases/
California Consumer Privacy Act, Cal. Civ. Code Sections 1798.100 et seq., as amended by the California Privacy Rights Act and implementing regulations, effective January 1, 2026.
Clyde and Co. ‘Latest Verdict Against Autonomous Vehicle Manufacturer Will Embolden Plaintiffs with Similar Claims.’ September 2025. https://www.clydeco.com/en/insights/2025/09/ruling-against-av-maker-may-boost-similar-lawsuits
Covington and Burling LLP. ‘NHTSA Proposes New Autonomous Vehicle Program.’ January 2025. https://www.cov.com/en/news-and-insights/insights/2025/01/nhtsa-proposes-new-autonomous-vehicle-program
Defend Trade Secrets Act, 18 U.S.C. Section 1836 (2016).
Eno Center for Transportation. ‘2025 Autonomous Vehicles Federal Policy Wrapped.’ December 2025. https://enotrans.org/article/2025-autonomous-vehicles-federal-policy-wrapped/
EU Artificial Intelligence Act, Regulation (EU) 2024/1689, Official Journal of the European Union, August 1, 2024.
Federal Register. ‘ADS-Equipped Vehicle Safety, Transparency, and Evaluation Program.’ 90 FR 4130, January 15, 2025. https://www.federalregister.gov/documents/2025/01/15/2024-30854/ads-equipped-vehicle-safety-transparency-and-evaluation-program
Federal Register. ‘Framework for Artificial Intelligence Diffusion.’ 90 FR 4608, January 15, 2025. https://www.federalregister.gov/documents/2025/01/15/2025-00636/framework-for-artificial-intelligence-diffusion
Food and Drug Administration. ‘2019 Proposed Regulatory Framework for Modifications to AI/ML-Based Software as a Medical Device.’ April 2, 2019.
Food and Drug Administration. Quality Management System Regulation, 21 C.F.R. Part 820, Final Rule, 89 FR 70396, September 2024, effective February 2, 2026.
Grand View Research. ‘Edge AI Market Size, Share and Trends Analysis Report.’ 2025. https://www.grandviewresearch.com/industry-analysis/edge-ai-market-report
Health Insurance Portability and Accountability Act, 45 C.F.R. Parts 160 and 164.
Illinois Biometric Information Privacy Act, 740 ILCS 14/1 (2008).
Innolitics. ‘2025 Year in Review: AI and ML Medical Device 510(k) Clearances.’ December 2025. https://innolitics.com/articles/year-in-review-ai-ml-medical-device-k-clearances/
International Traffic in Arms Regulations, 22 C.F.R. Parts 120 to 130.
Mondaq. ‘ITAR and AI-Enabled Defense Technologies: Autonomous Systems, Targeting Algorithms, and the New Export Control Frontier.’ 2025. https://www.mondaq.com/unitedstates/new-technology/1776516/itar-ai-enabled-defense-technologies-autonomous-systems-targeting-algorithms-and-the-new-export-control-frontier
Occupational Safety and Health Act, 29 U.S.C. Section 654.
O’Melveny. ‘2026 Data Security and Privacy Compliance Checklist.’ January 2026. https://www.omm.com/insights/alerts-publications/2026-data-security-and-privacy-compliance-checklist-key-us-state-law-updates-ai-rules-coppa-changes-and-global-data-protection-risks/
PatentPC. ‘Edge AI: Protecting Intellectual Property in On-Device Machine Learning.’ 2025. https://patentpc.com/blog/edge-ai-protecting-intellectual-property-in-on-device-machine-learning
Restatement (Third) of Torts: Products Liability (American Law Institute, 1998).
SAE International Standard J3016: Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles (2021).
Texas Business and Commerce Code Chapter 503, Capture or Use of Biometric Identifier.
Interested in analysis about the intersection of tech, policy and the law? Check out my Substack channel. https://theinnovationattorney.substack.com/