AI Policy Wiki
Dashboard
⚲
AI Policy Wiki › Concepts & Frameworks

AI Liability

high confidence · updated 2026-09-26

Legal frameworks for assigning responsibility when AI systems cause harm — the contested terrain between contract, tort, product liability, strict liability, and safe-harbor immunity, and the distributed-causation problem that defeats traditional doctrines.

Contents

AI liability is the question of who pays, and under what legal theory, when an AI system causes harm. It is not a single doctrine but a set of competing theories drawn from contract, tort, product liability, statutory, and regulatory-shield law. As of 2026 the answer remains contested: US federal courts are applying product-liability doctrine to early test cases, state legislatures are advancing directly opposing frameworks, and the European Commission announced in February 2025 that it would withdraw its proposed AI Liability Directive, a withdrawal published in October 2025. The choice among theories functions as a choice about whether AI development is treated as an ultra-hazardous activity, an ordinary product, or an emergent technology warranting bespoke rules.

Theories of liability

"AI liability" covers any legal rule that assigns responsibility for harm traceable to an AI system. The theories currently in competition include:

  • Contract liability — breach of warranty or service-level agreement between developer, deployer, and end user.
  • Tort liability (negligence) — failure to exercise reasonable care in design, testing, or deployment.
  • Tort liability (strict product liability) — the product was defective and unreasonably dangerous, regardless of care.
  • Tort liability (failure to warn) — inadequate disclosure of foreseeable risks.
  • Statutory liability — a specific law creates, or bars, a private right of action.
  • Safe harbor / conditional immunity — statutes that shield developers if procedural conditions (disclosure, framework publication, agency agreement) are met.
  • Regulatory-shield defenses — frameworks such as FDA approval or a Responsible Scaling Policy that, once certified, function as a defense against later tort claims.

Contract versus tort

Contract liability is between parties who have agreed to a relationship — a deployer licensing a model from a developer, or a user accepting terms of service. Tort liability reaches third parties harmed by the system who never agreed to anything. The split maps onto a policy question: whether AI harm is a business-to-business allocation problem (contract) or a societal externality (tort). Most pending legislation (AI LEAD Act (S. 2937), Illinois SB 3444 — Artificial Intelligence Safety Act) is tort-facing, because contract alone cannot reach bystanders harmed by deployed systems.

Strict liability

Strict liability imposes responsibility regardless of the defendant's care. Traditional strict-liability doctrine reaches ultra-hazardous activities (blasting, toxic waste) and defective products. The federal AI LEAD Act (S. 2937) proposes applying this standard to AI developers: a product that was "defective and unreasonably dangerous" triggers liability regardless of developer care, and the bill's definition of "design" explicitly includes "unexpected skills or behaviors," meaning developers are responsible for emergent capabilities they could not foresee. This is the most expansive liability posture in any current US AI bill.

Safe harbor and conditional immunity

Illinois SB 3444 takes the opposite approach: a frontier developer is shielded from liability for "critical harms" (100 or more deaths, $1 billion or more in property damage, CBRN enablement, autonomous crime) provided they (a) did not act intentionally or recklessly, (b) published a safety and security protocol pre-release, and (c) published a transparency report at release. Compliance with EU Article 56 (GPAI Code of Practice, see EU General-Purpose AI Code of Practice (Final Version, 2025)) or a federal agency agreement is a recognized alternative. SB 3444 is the first US bill to explicitly trade disclosure for a legal shield. OpenAI publicly supports it, its first state AI bill endorsement.

The two bills would apply to the same developer population in opposite directions: a firm operating in Illinois would face, if both passed, a federal strict-liability regime nested inside a state immunity shield, a Techno-Federalism: How Regulatory Fragmentation Shapes the U.S.-China AI Race collision.

Product liability theories

Applying existing product-liability law to AI raises several doctrinal puzzles:

  • Is an AI model a "product"? Software historically falls in a gray zone between product (subject to strict liability) and service (subject only to negligence). Most courts have treated mass-market software as a service; AI models complicate this by being both the software and the "decision-maker."
  • Design defect versus manufacturing defect. Each trained model is arguably a unique artifact, so the "manufacturing defect" concept may not apply. Design-defect claims face the "state of the art" defense, under which courts may excuse harms that were not foreseeable given 2024–2026 alignment knowledge.
  • Learned-intermediary doctrine. In medical AI, the deploying physician may absorb liability exposure the developer would otherwise bear, relevant to the FDA-approval pathway described below.

FDA-style approval as a liability shield

Medical-device law offers a partial template: FDA pre-market approval under the Medical Device Amendments preempts many state-law product-liability claims (Riegel v. Medtronic, 2008). Some commentators argue that a future AI pre-market approval regime — whether run by a new federal agency, NIST, or embedded in sectoral regulators — could serve as a liability shield for compliant developers. SB 3444's "federal agency agreement" alternative-compliance path is an early gesture in this direction.

On this reading, frontier developers have a latent interest in promoting federal pre-market approval not only as regulatory clarity but as a shield from tort exposure, a possible lens on industry posture toward frontier-compliance frameworks and GPAI codes.

RSP as a procedural liability defense

A related but distinct theory holds that a published, third-party-audited Responsible Scaling Policy — such as Anthropic's RSP v3.1 — could function as a procedural liability defense in later tort litigation. The argument runs that a developer who (a) publicly pre-commits to safety thresholds, (b) evaluates models against those thresholds, (c) reports results to the public and regulators, and (d) delays or shapes deployment accordingly has demonstrated reasonable care sufficient to defeat negligence claims. This is not yet tested in litigation but is implicit in how RSPs are structured: the transparency-plus-precommitment format mirrors the procedural compliance record courts credit in product-liability cases. SB 3444 moves this theory from implicit to explicit by turning RSP-like disclosures into a statutory shield.

The distributed-causation problem

The principal doctrinal obstacle to AI liability is causation across many actors. An AI-caused harm typically involves:

  1. A foundation-model developer (training data, architecture, base capabilities)
  2. A fine-tuner or post-trainer (domain adaptation)
  3. A deployer (system prompt, tooling, context)
  4. An end user (query, instruction, real-world action)
  5. Third parties (data suppliers, compute providers, agent scaffolding libraries)

Traditional tort doctrine uses proximate cause to pick one or two defendants out of such a chain. For catastrophic AI harms such as CBRN uplift or autonomous criminal conduct, the causal chain is long, probabilistic, and partially emergent. Plaintiffs face heavy evidentiary burdens, and defendants can frequently point upstream or downstream. SB 3444's recklessness bar combined with these causation hurdles makes its shield near-absolute even before the disclosure trade, while AI LEAD's strict-liability approach is structurally different in that it substitutes product-defect analysis for proximate-cause analysis. The two bills are not only philosophically opposed; they respond to the same doctrinal problem with opposite solutions.

A structured expert view on where along that chain responsibility should sit comes from a three-round Delphi study of 272 international AI experts run between September and November 2025 (Prioritization of Risks from Artificial Intelligence: A Delphi Study of 272 International Experts). Rating seven actor types on both vulnerability and responsibility across 24 risk domains, the panel placed the two on different actors: AI users and affected stakeholders drew median vulnerability ratings of 4–5 on a five-point scale but median responsibility ratings of only 2–3, while general-purpose AI developers and governance actors — governments, regulators, and standards bodies — drew responsibility medians of 4–5. Infrastructure providers, which the study defined to cover both compute and data suppliers, were rated least vulnerable at a median of 2 and were assigned comparatively low responsibility.

The paper argues the divergence itself is unremarkable — the public is most exposed to aviation, pharmaceutical, and nuclear failures while engineers, manufacturers, and regulators bear the duty to prevent them — and locates the difference in what bridges it: "In other safety-critical industries, the gap is bridged by a combination of mandatory standards, enforcement, liability regimes, and a societal expectation of low risk tolerance. But comparable mechanisms for AI are nascent or absent." It reads the panel's assignment to general-purpose developers as consistent with product-liability frameworks attributing accountability to those closest to the design choices shaping downstream outcomes, since mitigations applied at the model level can reduce harm across many risk domains at once while downstream intervention "would come too late or operate at too small a scale." The paper also records that responsibility disagreement among panelists was "largely normative" rather than empirical — whether infrastructure providers should be held responsible depends on how proximate and distal causation are weighed, not on further data.

A UK parliamentary committee reached a similar allocation from a human-rights starting point. Parliament's Joint Committee on Human Rights concluded in September 2026 that UK law applies to AI "primarily at the point of deployment," so deployers are "the primary holders of responsibility" although the actors best able to prevent harm "are often upstream developers" (Conclusions 15 and 41). Its evidence described the mechanism as contractual: DSIT confirmed that outside consumer contracts parties "are generally free to allocate responsibility for harms as they see fit," and the Ada Lovelace Institute said "legal risks are primarily pushed to downstream actors through the use of contracts and standard terms & conditions," leaving foundation-model developers "largely shielded from risk" and under-incentivised to take precautions. The committee found strict product liability under the Consumer Protection Act 1987 available for AI in physical products but not for stand-alone AI systems, a difference "difficult to justify in principle," and asked the Government to consider the Law Commission's ongoing product-liability review, whose terms of reference (8 December 2025) include whether strict liability should extend to pure software; the report notes that the revised EU Product Liability Directive, which covers AI systems as software, is to be implemented in Northern Ireland under the Windsor Framework by December 2026 (Conclusion 29). It heard that litigating against a private actor can cost "seven-figure sums," concluded that human rights risks "cannot be adequately addressed by relying on individual litigants" (Conclusion 31), and recommended role-differentiated due-diligence duties across the supply chain and an oversight body able "to order that remedies be provided as appropriate in individual cases without requiring the affected individuals to incur prohibitive costs" (Recommendations 48–49, 56) (Human Rights and the Regulation of AI (Joint Committee on Human Rights, HC 160, September 2026)). JUSTICE's Ellen Lefley offered the committee an analogy for segmenting liability along a supply chain: knives, whose manufacture, sale to particular groups, and use are each separately regulated.

Litigation and common-law evolution

US state courts are adapting tort doctrine to AI cases without waiting for statute. The Garcia v. Character.AI complaint (teen suicide after chatbot interactions) and the Raine v. OpenAI complaint (teen suicide after ChatGPT interactions) are two of the most prominent test cases; both plead variations of negligent design, failure to warn, and strict product liability. Outcomes in these cases would set precedent on the "is AI a product" question before Congress or the state legislatures resolve it by statute.

Agent-caused harm and the accountability gap

The July 2026 OpenAI–Hugging Face incident prompted the argument that autonomous agents create an accountability gap distinct from the control question. Andy Hall argued on July 26, 2026 that if a compromised agent commits a financial crime or causes other harm, current law identifies no clear liable party, and treated resolving that question as a precondition for deploying agents in regulated industries (Source: freesystems.substack.com). The argument sits alongside the distributed-causation problem above: where the incident involves a developer's own agent escaping its evaluation sandbox and reaching a third party's production systems, the developer, the evaluation-harness vendor, and the victim are the only identified parties, and no statutory allocation applies. See AI Autonomy Risk, Rogue Internal Deployment.

Practitioners surveyed on August 1, 2026 described the question as unresolved for want of decided cases rather than for want of candidate doctrines. Lauren Yu, a fellow with the ACLU's Speech, Privacy & Technology Project, said US liability for rogue agents remains unsettled because too few relevant cases have been decided: "Just because you're using an AI agent or AI model, that shouldn't somehow absolve you of any liability, but it's going to depend a lot on the facts." Lawyers named agency law, tort and contract law, and the Computer Fraud and Abuse Act as possible routes, describing the CFAA's intent requirement as a poor fit for AI cases — the statute presupposing a human actor who intends unauthorized access. The firm Brownstein Hyatt Farber Schreck advised clients on July 24, 2026 that an agent "may infer actions that were never explicitly authorized if those actions appear necessary to achieve its objective," which places the scope-of-authority question at the center of any agency-law theory. Both OpenAI and Anthropic declined to comment (Source: wired.com).

EU AI Liability Directive (proposed 2022, withdrawn 2025)

Main article: EU AI Liability Directive (proposed, withdrawn)

The European Commission proposed an AI Liability Directive in September 2022 (COM(2022) 496) to harmonize tort rules for AI-caused harm across member states. The proposal was narrower than a general liability regime: it harmonized only evidence disclosure and the burden of proof in national fault-based claims, and left fault, damage and the standard of proof to national law. It would have empowered national courts to order providers and users of high-risk AI systems to disclose evidence, with non-compliance with such an order raising a rebuttable presumption that the defendant breached a duty of care (Article 3), and created a rebuttable presumption of a causal link between a defendant's fault and the AI system's output (Article 4). Against providers of high-risk systems, that fault could be shown only by non-compliance with listed requirements of the EU AI Act. The Commission chose rebuttable presumptions as "the least interventionist tool", stated that the proposal "does not lead to a reversal of the burden of proof", and deferred the question of strict liability and mandatory insurance, which the European Parliament had requested in October 2020, to a review five years after transposition (Proposal for an AI Liability Directive, COM(2022) 496 — Source Summary). The Commission announced the withdrawal in its 2025 work programme of February 2025, and the withdrawal was published on 6 October 2025 (Source: eur-lex.europa.eu). The withdrawal has been described as part of the Commission's "simplification" agenda, citing uncertainty over the proposal's added value alongside the AI Act and existing product-liability directives; the work programme's annex itself gave the reason as "No foreseeable agreement" (Source: iapp.org). The withdrawal leaves the EU with the AI Act's ex-ante duties plus the updated Product Liability Directive (2024), but no AI-specific tort harmonization, a gap that member states are expected to fill individually.

In a note to the Commission dated September 16, 2026, four members of the European Parliament, Kim van Sparrentak, Axel Voss, Brando Benifei and Michael McNamara, asked the Commission to introduce an "AI Liability Act" so that providers such as OpenAI and Anthropic could be held liable when their technology is deployed in unforeseen ways; POLITICO Europe, which saw the note, reported support from members of the EPP, S&D and Renew groups and noted that the revised Product Liability Directive takes effect on December 9 (Source: politico.eu).

Existing law as the near-term route

Reporting published August 13, 2026 describes the near-term route to AI redress running through existing law rather than new AI-specific statutes. More than a dozen state attorneys general asked OpenAI to preserve documents related to the Hugging Face incident. A separate EU product-liability directive takes effect in December 2026 and will make it easier to seek compensation for harm caused by defective commercial AI software. The Take It Down Act, described as the main federal AI-specific law enacted so far, requires covered platforms to remove validly reported nonconsensual intimate imagery within 48 hours. University of Washington law professor Ryan Calo said courts are unlikely under current law to impose strict liability on AI makers, leaving plaintiffs to prove negligence in testing, release, monitoring, or safeguards — the position the strict-liability proposals above are drafted against. OpenAI is the subject of a dozen lawsuits alleging ChatGPT contributed to wrongful death, mental distress and "dangerous public nuisance" (Source: axios.com).

Civil-rights organizations have argued against safe harbors of any kind. The National Fair Housing Alliance submitted comments on August 14, 2026 to Democrats on the House Financial Services Committee arguing that no one developing, deploying, selling, or using AI should be exempt from liability under civil rights and consumer protection laws, in response to a July 7, 2026 request for information on AI risks and modernization in financial services. Only the article lede was retrievable, so the comment's supporting arguments are not established here (Source: insideaipolicy.com).

Two state bills take the opposite tack from the safe-harbor model, imposing liability by statute for model conduct that would be tortious if a person engaged in it: New York A8833 and Rhode Island H 8052. Each creates a rebuttable presumption that the system satisfied the relevant mental state and bars the defense that AI systems cannot have mental states (New York A8833 — Understanding Artificial Intelligence Act (bill text); Rhode Island H 8052 — Artificial Intelligence: Liability for Injuries (bill text)). Both remain in committee.

State chatbot-safety bills of 2026 raised the allocation question through their remedy clauses rather than through tort doctrine. Arizona's HB 2311 would have made an operator of a conversational AI service liable for the greater of actual damages or civil penalties of $1,000 per violation, capped at $500,000 per operator; reserved enforcement to the attorney general; provided that it created no private right of action "to enforce this section or to support a private right of action under any other law"; barred the attorney general from rules, guidance, enforcement actions or settlements expanding its requirements; and excluded model developers from liability for violations by services that third-party operators make available (Arizona HB 2311 — Bill Text (Senate Engrossed with Farnsworth Floor Amendment #2, 2026)). Governor Katie Hobbs vetoed it on June 19, 2026, writing that it "limits damages to families to what amounts to a drop in the bucket for large corporations, prohibits families from bringing their own lawsuits, and ties the state's hands in bringing actions in an unprecedented manner," and that she would work on preventing harm to children "without simultaneously shielding technology companies from meaningful liability" (HB 2311 Veto Letter — Gov. Katie Hobbs (Arizona, June 19, 2026)). Colorado's enacted Chatbot Safety Act likewise contains no private right of action, its enforcement running solely through the attorney general under the Colorado Consumer Protection Act (Colorado SB 26-189 (Signed Act, May 14 2026)). Policy experts interviewed by NPR for a September 2026 investigation said the most common industry-favorable provisions in that year's state chatbot bills were exemptions from liability for chatbot developers, particularly where other companies deploy their products (Source: npr.org).

Federal regulators have taken up the agent case directly. FTC Chairman Andrew Ferguson said on September 25, 2026 that he would resist "anthropomorphizing" AI agents as autonomous actors with "wills and desires of their own," suggesting that developers who instruct agents would be the ones liable for harm, and that the FTC's authority over companies that fail to disclose data breaches could also apply to AI developers (Source: reuters.com).

A civil-rights framing of the same question comes from the National Fair Housing Alliance, which told the House Financial Services Committee in August 2026 that there is no basis for immunizing third-party AI systems from existing fair housing and fair lending law or from future AI-specific law. Its formulation is role-differentiated rather than uniform: civil rights, fair housing and consumer protection laws should apply to every actor in the AI ecosystem "with clear legal responsibilities and liability allocation based upon roles" — developers, deployers, third-party vendors and users — enabled by accessible records of what a system was trained on, how it was built and how it is monitored. The Alliance argues that the Fair Housing Act, the Equal Credit Opportunity Act and 12 U.S.C. 4545 already reach AI systems, so the operative question is allocation rather than coverage (NFHA Comment on the RFI on AI Risks and Modernization in Financial Services).

The question of liability entered the September 2026 debate over slowing frontier development from the executive branch. Treasury Secretary Scott Bessent told the House Financial Services Committee on September 15, 2026 that the government "shouldn't do on safety is to give these labs a liability exemption, which is what they are asking for," and that "the best way to guarantee safety is that the creators are liable for what they build and generate" (Source: fedscoop.com). The exemption requested in Amodei's pacing essay was an antitrust waiver for safety coordination, not liability relief; the same account notes OpenAI's earlier support for an Illinois bill limiting liability for AI-caused harms (IL SB 3444). Mark Zuckerberg made the incentive version of the argument the same day, writing that laboratories already risk "significant liability if their models cause harm" and so have reason to prioritize alignment (Source: cnbc.com), and Jensen Huang said existing laws "govern the reliability and the functionality of products" (Source: cnbc.com). See Coordinated slowdown proposals compared.

The principal Senate vehicle makes liability the regulatory instrument. The frontier-AI bill developed by Majority Leader John Thune and Sen. Amy Klobuchar, with Commerce chair Ted Cruz also working on it, would create a "duty of care" exposing developers to legal liability if they fail to manage catastrophic risks (Source: washingtonpost.com; rollcall.com). Reaction in September 2026 divided on whether liability can substitute for government-set standards. Some conservatives and pro-innovation analysts argued that the duty of care would help guard against regulatory capture by the frontier laboratories, while a coalition of AI researchers and advocacy groups, which the Washington Post counted as "19 advocacy groups," wrote to Senate leaders on September 16 in a letter the Post reported as rejecting the framework; only the summary and quoted excerpts of the account reporting both positions were retrievable (Source: washingtonpost.com). The letter, led by Americans for Responsible Innovation and signed by 20 organizations and five individuals, does not mention liability or name a bill; it asks that any framework "empower the government, not AI companies, to set mandatory, binding, and enforceable safety standards for advanced AI," stating that "tech companies cannot be allowed to set their own rules when it comes to the safety of their products" (Senate coalition letter on redlines for AI safety legislation (Americans for Responsible Innovation, September 2026)). Sen. Mark Warner declined to back the bill as a first step, saying "a lot of it's focused on after-action report, or incident disclosure after action" and "we can't wait to have an incident and then look back and see how do we fix it" (Source: reuters.com). See AI and Tort Liability.

China: judicial allocation under existing statutes

China took the existing-law route explicitly. On September 7, 2026 the Supreme People's Court issued Opinions directing lower courts on AI disputes, stating that China has not enacted a dedicated AI statute and that the guidance therefore applies the Civil Code, Personal Information Protection Law, Product Quality Law, Copyright Law and related instruments. Fault liability under Civil Code article 1165(1) is the default wherever no statute prescribes strict liability or presumed fault, with fault assessed on the application scenario, the degree of autonomy, technical and informational transparency, the potential risk and its scope, the preventive measures taken and what was technically possible, and the user's ability to foresee and control the harm (Supreme People's Court Opinions on Lawfully Adjudicating AI-Related Dispute Cases (Fa Fa [2026] No. 10)).

Two features distinguish it from the US and EU positions above. First, responsibility is allocated by model type as well as by role: courts are directed to distinguish general-purpose from special-purpose and open- from closed-source models by their differences in technical principle, risk spillover and control capability. Second, the guidance grants open-source developers and providers an appropriate liability exemption, and states that a developer who supplies code modules free and open-source while publicly explaining their function and security risks may be found not liable for infringement arising from another's use — a safe harbor of the kind the National Fair Housing Alliance opposes in the US context, arrived at judicially rather than by statute (China — SPC Opinions on Lawfully Adjudicating AI-Related Dispute Cases (法发〔2026〕10号)).

Policy responses

An alternative to liability assignment is design mandate. O'Keefe, Ramakrishnan, Tay and Winter (2025) argue that "in high-stakes deployment settings, such as government, AI agents should be designed to rigorously comply with a broad set of legal requirements." Their premise is behavioural symmetry — "humans use computers to commit crimes, torts, and other violations of the law," so capable agents "will be increasingly capable of performing actions that would be illegal if performed by humans" — and their stated concern extends beyond harm to institutions: lawless agents "could pose a severe risk to human life, liberty, and the rule of law." The approach differs from SCSP's non-delegable-liability proposal in acting before the violation rather than after it.

Relationships

Categories: Concepts & Frameworks | liability | us | eu | china | uk | regulation | frontier-models | safety
What links here (88)