Narrative Goldmine

AI Liability

39 min read

Properties

Type
  • Class
Resource
  • urn:ngm:class:ai-liability
Domain
  • artificial-intelligence
Status
  • stable
Maturity
  • established
Quality
  • 0.52
Is a
Has part
Requires
Enables
Generated
  • by: process:vault-migrate/1.0 · at: 2026-09-22T12:41:39.024503178Z
Contrasts with
Depends on
Implemented in layer
Implements
Legacy term id
  • AI-1087
Links
Page resource
  • urn:visionflow:page:08f8a95794397b82a254bfe32333f811212c444b34c53a572638102e400d0613
Related to
Standardized by
Supports
Uses

AI Liability is the body of legal doctrine, statutory regime, regulatory practice and emerging case law that allocates civil responsibility for personal injury, property damage, economic loss, discrimination, privacy invasion and consequential harm caused by artificial intelligence systems—encomp…

Semantic Classification

Content

## Compositional Relationships (Components)
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:hasPart ai:DutyOfCare))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:hasPart ai:BreachOfDuty))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:hasPart ai:Causation))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:hasPart ai:Foreseeability))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:hasPart ai:Damages))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:hasPart ai:StrictLiabilityStandard))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:hasPart ai:ConformityAssessment))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:hasPart ai:PostMarketMonitoring))

## Dependency Relationships
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:requires ai:IdentifiableLegalPerson))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:requires ai:CompensableHarm))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:requires ai:EvidentiaryRecord))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:requires ai:CausalLinkage))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:dependsOn ai:CommonLawSystem))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:dependsOn ai:StatutoryInterpretation))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:dependsOn ai:JudicialCapacity))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:dependsOn ai:ExpertEvidence))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:dependsOn ai:Precedent))

## Capability Relationships
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:enables ai:CompensationForVictims))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:enables ai:DeterrenceOfHarm))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:enables ai:RiskInternalisation))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:enables ai:InsurancePricing))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:enables ai:SafetyInvestment))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:supports ai:AISafety))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:supports ai:MarketSurveillance))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:supports ai:AlgorithmicAuditing))

## Implementation Relationships
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:implements ai:NegligenceStandard))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:implements ai:StrictLiability))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:implements ai:VicariousLiability))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:implements ai:AbnormallyDangerousActivityDoctrine))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:implements ai:ProductLiabilityDirective))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:implements ai:EUAIAct))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:uses ai:ForeseeabilityTest))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:uses ai:CaparoThreeStageTest))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:uses ai:DaubertStandard))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:uses ai:ReasonableCareStandard))

## Reduction Relationships
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:reduces ai:UncompensatedHarm))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:reduces ai:MoralHazard))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:reduces ai:InformationAsymmetry))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:reduces ai:SocialCostExternalisation))

## Association Relationships
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:relatedTo ai:AIEthics))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:relatedTo ai:AlgorithmicBias))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:relatedTo ai:FrontierAI))
SubClassOf(ai:AILiability
  ObjectSomeValuesFrom(ai:contrastsWith ai:CriminalLiability))

## Data Properties (Characteristics)
DataPropertyAssertion(ai:hasIdentifier ai:AILiability "AI-1087"^^xsd:string)
DataPropertyAssertion(ai:authorityScore ai:AILiability "0.87"^^xsd:decimal)
DataPropertyAssertion(ai:euAIActEntryIntoForce ai:AILiability "2024-08-01"^^xsd:date)
DataPropertyAssertion(ai:revisedPLDEntryIntoForce ai:AILiability "2024-12-09"^^xsd:date)
DataPropertyAssertion(ai:euAIActHighRiskPenaltyCeilingEUR ai:AILiability "35000000"^^xsd:integer)
DataPropertyAssertion(ai:euAIActHighRiskPenaltyTurnoverPct ai:AILiability "0.07"^^xsd:decimal)
DataPropertyAssertion(ai:aiInsuranceMarketUSD2025 ai:AILiability "1200000000"^^xsd:integer)
DataPropertyAssertion(ai:riteAidFTCBanYears ai:AILiability "5"^^xsd:integer)

## Property Constraints
SubClassOf(ai:AILiability
  DataMinCardinality(1 ai:hasJurisdiction xsd:string))
SubClassOf(ai:AILiability
  DataMinCardinality(1 ai:hasLegalBasis xsd:string))
SubClassOf(ai:AILiability
  DataAllValuesFrom(ai:hasResponsibleParty xsd:string))

## Annotations
AnnotationAssertion(rdfs:label ai:AILiability "AI Liability"@en)
AnnotationAssertion(rdfs:comment ai:AILiability "Body of legal doctrine, statutory regime and regulatory practice allocating civil responsibility for harm caused by AI systems, encompassing the application of common-law negligence and strict liability, statutory product-liability regimes including the EU revised Product Liability Directive (EU) 2024/2853 and Consumer Protection Act 1987, the EU AI Act conformity-assessment and market-surveillance framework, the doctrine of abnormally dangerous activity proposed for frontier AI, vicarious liability for AI agents, sector-specific regimes for autonomous vehicles (NHTSA, ASDE) and medical devices (FDA, MHRA SaMD), and an emerging insurance market spanning Munich Re aiSure, Lloyd's syndicates and InsurTech carriers Vouch/Armilla."@en)
AnnotationAssertion(dcterms:identifier ai:AILiability "AI-1087"^^xsd:string)
AnnotationAssertion(dcterms:subject ai:AILiability "Tort Law, Product Liability, AI Governance, Regulation, Insurance, Algorithmic Accountability"@en)

About AI Liability

  • AI Liability is the legal architecture for assigning civil responsibility when Artificial Intelligence systems cause harm. It sits at the intersection of Tort Law (negligence, strict liability, vicarious liability), Product Liability (the UK Consumer Protection Act 1987, the EU revised Product Liability Directive (EU) 2024/2853, the US Restatement (Second) of Torts § 402A), administrative regulation (the EU AI Act Regulatory Instrument conformity-assessment regime, the FDA AI/ML SaMD framework, NHTSA’s Automated Vehicles Policy, the MHRA Software and AI as Medical Device programme, the FTC’s deceptive-practices and Section 5 enforcement) and an increasingly active insurance market (Munich Re aiSure, Lloyd’s syndicate coverage, Vouch and Armilla AI). It addresses the central question facing legal systems in 2026: when an autonomous, opaque and adaptive AI system produces an output that causes injury, who bears the loss—the developer, the integrator, the deployer, the user, or some combination?
  • The unique characteristics of advanced AI—autonomy (capacity to produce actions not directly programmed), opacity (the black box problem of deep neural networks), adaptivity (learning and behavioural drift after deployment), and emergence (capabilities arising unexpectedly during scale-up)—strain the foundational doctrines of common law tort, which presuppose human agency, foreseeability, and traceable causation. These features force the law into one of three responses: (1) adapt existing doctrines through clever procedural devices (rebuttable presumptions of causation as proposed in the now-withdrawn AI Liability Directive AILD), (2) codify new statutory regimes specific to AI (the EU AI Act Regulatory Instrument, Colorado AI Act SB 24-205, Rhode Island S0358), or (3) shift to no-fault regimes for the highest-risk applications (treating frontier-model development as an Abnormally Dangerous Activity or mandating compulsory insurance modelled on the Price-Anderson Act for civil nuclear power).
  • Three trans-Atlantic developments dominate the 2025-2026 landscape. First, the EU AI Act (Regulation (EU) 2024/1689) became the world’s first comprehensive horizontal AI statute, with obligations for general-purpose AI models commencing 2 August 2025 and high-risk obligations applying 2 August 2026. Second, the revised Product Liability Directive (EU) 2024/2853, entering into force on 9 December 2024 with Member-State transposition by 9 December 2026, expressly redefines “product” to include software and AI systems, introduces a presumption of defectiveness when claimants face “excessive difficulties” proving technical complexity, and abolishes the €500 deductible. Third, the AI Liability Directive (AILD COM(2022) 496) was withdrawn by the European Commission on 11 February 2025 as part of the 2025 Work Programme reset, on grounds that “no foreseeable agreement” existed—a significant gap that leaves causation rules for AI-related harms to fragmented Member-State tort regimes and the residual force of the revised PLD.
  • Across the Atlantic, the United States is following a fragmented “bottom-up” path. There is no comprehensive federal AI liability statute. Instead, individual states have legislated targeted regimes (Colorado AI Act SB 24-205 enacted May 2024 covering consequential decisions, effective February 2026; New York City Local Law 144 of 2021 mandating bias audits for AEDTs; Texas Responsible AI Governance Act HB 1709 advancing through 2025), federal agencies have exercised existing authorities (FTC Section 5 actions against Rite Aid December 2023 for facial-recognition harms, the Amazon Alexa COPPA settlement of $25M in 2023, the Operation AI Comply sweep September 2024), and the courts have begun reshaping product-liability doctrine to treat AI software as a “product” subject to strict liability (the Character.AI ruling in Garcia v. Character Technologies M.D. Fla. May 2025).
  • In the United Kingdom, the pro-innovation principles-based framework of the March 2023 White Paper survives but is under sustained Parliamentary and academic pressure. The House of Commons Science, Innovation and Technology Committee called for AI-specific legislation in its August 2024 final report; Lord Holmes’ Artificial Intelligence (Regulation) Bill [HL] was reintroduced in March 2025; and the Law Commission continues scoping work on autonomous-systems liability following the Automated Vehicles Act 2024 (Royal Assent 20 May 2024) which implements the joint Law Commission/Scottish Law Commission proposals on the Authorised Self-Driving Entity (ASDE) framework. The UK AI Safety Institute was renamed the AI Security Institute in February 2025, narrowing its remit to security-related model evaluations.

Components and Architecture

  • The legal architecture of AI liability decomposes into eight interlocking components.

1. Cause of Action

2. Standard of Care

  • Negligence claims turn on the standard of care expected of a reasonable developer, integrator or deployer. Standards are crystallising through several channels:
    • Technical standards bodies. ISO/IEC 42001 AI Management System published December 2023, ISO/IEC 23894 AI risk management, ISO/IEC TR 24028 AI trustworthiness, the NIST AI Risk Management Framework AI RMF 1.0 published January 2023 with the Generative AI Profile (NIST AI 600-1) released July 2024
    • Sector regulator guidance. The ICO AI auditing framework and Guidance on AI and Data Protection (updated 2023), the Bank of England/PRA Supervisory Statement SS1/23 on Model Risk Management Principles, the FCA’s AI Update (October 2024)
    • Codes of practice. The EU AI Act Code of Practice for General-Purpose AI Models published May 2025 covering transparency, copyright and safety/security obligations for GPAI models with systemic risk above 10²⁵ FLOPs
    • Professional rules. ABA Formal Opinion 512 on generative AI in legal practice (July 2024), the Bar Standards Board guidance (December 2023), the SRA Risk Outlook (2024)

3. Causation

  • Establishing the causal link between an AI system’s behaviour and the claimant’s harm is the most difficult evidentiary task in AI litigation.
    • The black box problem makes it impossible in many cases to trace the specific computational pathway leading to a particular output
    • The withdrawn AILD proposed a rebuttable presumption of causation when the claimant demonstrated the defendant’s breach of a duty of care relevant to the harm
    • This principle survives in scholarly proposals and continues to influence Member-State courts
    • The revised PLD (EU) 2024/2853 Article 10 introduces a parallel presumption of defectiveness when the claimant faces “excessive difficulties” in proving technical complexity
    • Article 10 shifts the evidentiary burden to the producer once the claimant has shown likely defect and that defect-caused harm
    • Common-law alternatives include the “material contribution” test in Bonnington Castings v. Wardlaw [1956] AC 613 and the “but for” test of factual causation

4. Damages

  • Compensable damages under traditional tort law include personal injury, property damage, pure economic loss (limited in many jurisdictions), and—in narrowly defined circumstances—psychiatric injury and reputational harm.
    • The revised PLD expands the categories of recoverable damage to include destruction or corruption of data not used for professional purposes and medically recognised psychological harm
    • Punitive Damages (called exemplary damages in the UK) are available in the US under state law subject to constitutional due-process ceilings established in BMW v. Gore 1996 and State Farm v. Campbell 2003
    • In the UK, exemplary damages are confined by Rookes v Barnard 1964 to three narrow categories: oppressive government conduct, profit-calculated wrongdoing, statutory authorisation
    • Gabriel Weil’s “warning shot” thesis proposes calculating punitive damages on near-miss incidents to internalise the probability-weighted cost of unrealised catastrophic risk
    • Class-action and group-litigation damages are increasingly significant: MDL 3047 (social-media addiction), the New York Times v. OpenAI mass-copyright theory, the BC Court of Appeal Air Canada precedent

5. Defendants and Liability Chains

  • AI value chains often involve multiple parties:
    • Foundation-model provider (OpenAI, Anthropic, Google DeepMind, Meta, Mistral, xAI, DeepSeek)
    • Fine-tuner or integrator adapting the base model for a specific use case
    • Deployer embedding the model in a product or service offered to end-users
    • End-user or affected third party
  • Allocation rules vary by regime:
    • The EU AI Act allocates obligations along this chain to “providers” and “deployers,” with explicit downstream-modification rules (Article 25)
    • The revised PLD extends liability to manufacturers, importers, authorised representatives, fulfilment service providers and, in some cases, distributors
    • The Rhode Island S0358 bill places primary strict liability on developers of covered models trained above 10²⁶ FLOPs or costing >$100M
    • Litigation strategy in foundation-model cases (Andersen v. Stability AI, Getty v. Stability AI, the New York Times v. OpenAI) involves complex joinder of multiple defendants across the development chain

6. Defences

  • Defendants invoke several categories of defence:
    • Statutory defences. The development-risks/state-of-the-art defence under CPA 1987 section 4(1)(e), the manufacturer-aligned defences under the revised PLD Article 11 (defect did not exist when placed on market, compliance with mandatory regulations)
    • Common-law defences. Contributory negligence, volenti non fit injuria (consent to risk), intervening cause (novus actus interveniens), illegality (ex turpi causa)
    • Constitutional defences. In the US, Section 230 of the Communications Decency Act for platform liability—though its application to AI outputs is contested in cases including Lemmon v. Snap 9th Cir. 2021 and the Character.AI litigation
    • First Amendment defences in defamation-by-AI cases such as Walters v. OpenAI Georgia 2023 (dismissed May 2024)
    • Open-source defences. Apache 2.0/MIT licence disclaimer of warranties and limitation of liability—though increasingly contested for commercial deployers

7. Insurance and Risk Transfer

  • Insurance has become a quasi-regulatory mechanism for AI risk:
    • Munich Re aiSure launched 2018 as a performance-warranty product backing third-party AI vendors, expanded 2024 for generative AI
    • Lloyd’s of London syndicates (Tokio Marine Kiln, Beazley, Chaucer, Hiscox) write affirmative AI coverage including IP indemnity for foundation-model users
    • InsurTech specialists Vouch and Armilla AI offer dedicated AI policies for fintech and SaaS firms
    • Cyber carriers including Coalition, At-Bay and CFC have introduced AI-specific endorsements covering prompt-injection, data-poisoning and model-extraction attacks
    • AIG and Zurich market AI-risk addenda to professional indemnity and E&O policies
    • The market is small but growing rapidly—estimated at 4-6B by 2030
    • Reinsurance interest from Swiss Re, Hannover Re and SCOR underwrites the primary market

8. Procedural Architecture

  • AI cases produce unique procedural challenges:
    • Expert evidence on neural network behaviour governed in the US by Daubert v. Merrell Dow Pharmaceuticals 1993 standards, in the UK by the Civil Procedure Rules Part 35
    • Disclosure of training data and model weights (the Andersen v. Stability AI plaintiffs’ efforts to subpoena the LAION-5B dataset; the Getty UK litigation’s e-discovery battles over model checkpoints)
    • Expert assessors under CPR 35.15 and 35.7 single joint experts
    • Judicial education programmes (the Federal Judicial Center in the US, the National Courts and Sciences Institute, the Judicial College in the UK, the European Judicial Training Network)
    • Specialist courts and chambers including the Chancery Division’s Intellectual Property List, the Technology and Construction Court, and proposed AI-specialised divisions in Germany and France

Use Cases / Major Families

  • AI liability divides into seven major families distinguished by harm type, regulatory regime and characteristic defendants.

Autonomous Vehicle Liability

  • The most mature AI-liability domain.
    • In the US, NHTSA regulates motor-vehicle safety under FMVSS and exercises recall authority
    • Standing General Order 2021-01 (June 2021, amended April 2023) requires manufacturers and operators of Level 2+ ADAS and Level 3-5 ADS vehicles to report crashes within 24 hours
    • This data underpinned NHTSA’s December 2023 recall of 2 million Tesla vehicles for Autopilot driver-engagement defects (recall 23V-838)
    • Continuing investigations target Cruise (operations suspended October 2023 after the San Francisco dragging incident) and Waymo
    • In the UK, the Automated Vehicles Act 2024 establishes the Authorised Self-Driving Entity (ASDE) framework
    • The In-Use Regulatory Scheme is administered jointly by DfT, DVSA and the new regulator under Statutory Instruments to be made 2025-2027
    • The Authorised Self-Driving Entity—not the user-in-charge—bears criminal and civil responsibility for the vehicle’s driving behaviour while in self-driving mode
    • Notable cases include Banner v. Tesla (Florida state, jury verdict August 2023 finding Tesla 33% at fault in fatal Autopilot crash)
    • The Mercedes Drive Pilot Level 3 deployment in California and Nevada (2023-2024) and the BMW Personal Pilot L3 (Germany 2024) illustrate the OEM-led Level 3 approach

Medical AI Liability

  • Governed by the FDA AI/ML Software as a Medical Device (SaMD) regime in the US and the MHRA Software and AI as Medical Device programme in the UK.
    • In the EU, the MDR 2017/745 and IVDR 2017/746 apply, with EU AI Act high-risk obligations layering on top
    • FDA’s January 2021 AI/ML SaMD Action Plan introduced the Predetermined Change Control Plan (PCCP) framework
    • PCCP was formalised in the December 2024 Final Guidance, permitting pre-authorised algorithm updates without 510(k) resubmission
    • As of mid-2025 the FDA had cleared over 1,000 AI/ML-enabled medical devices
    • The MHRA Software and AI as Medical Device Change Programme (announced 2021, Roadmap 2024) coordinates 11 work-packages including the AI Airlock regulatory sandbox for adaptive AI
    • Liability theories include traditional medical malpractice (where the AI is a tool used by a clinician), product liability against the device manufacturer, and emerging “negligent reliance” claims against clinicians who over-trust algorithmic outputs
    • Notable cases include the IDx-DR (diabetic-retinopathy autonomous diagnosis) regulatory framework and the litigation tail from the IBM Watson Oncology MD Anderson divestment

Employment AI Liability

  • Driven by anti-discrimination law and emerging algorithmic-accountability statutes.
    • EU AI Act Annex III classifies employment-related AI (recruitment, performance assessment, allocation of tasks) as high-risk
    • New York City Local Law 144 of 2021 (effective July 2023) mandates annual bias audits of Automated Employment Decision Tools (AEDTs) by an independent auditor and candidate notice
    • Colorado AI Act SB 24-205 (effective February 2026) requires consequential-decision developers and deployers to use reasonable care, conduct impact assessments and notify subjects
    • UK employment AI is governed by the Equality Act 2010, the UK GDPR (Article 22 on automated decision-making), and the ICO’s Guidance on AI and Data Protection
    • The EEOC’s May 2023 technical guidance on Title VII and AI provided the federal-US baseline
    • The OFCCP’s AI directive (2024) applies to federal contractors
    • Notable cases include Mobley v. Workday N.D. Cal. (filed 2023, certification of conditional collective May 2024) alleging Workday’s screening AI discriminates against older Black applicants

Generative AI Content Liability

  • A rapidly developing area covering copyright infringement in training data, output infringement, defamation by hallucination, and consumer-protection violations.
    • Andersen v. Stability AI N.D. Cal. (filed January 2023, motion-to-dismiss ruling August 2024 narrowing but preserving direct-infringement and DMCA s. 1202(b) claims)
    • Getty Images v. Stability AI UK High Court (trial commenced June 2025, judgment awaited late 2025 on copyright infringement and trade-mark dilution; parallel US action in D. Del. continues)
    • New York Times Co. v. Microsoft & OpenAI S.D.N.Y. (filed December 2023 alleging mass copying of Times articles for GPT training)
    • Tremblay/Silverman/Chabon v. OpenAI consolidated authors’ MDL
    • Concord Music v. Anthropic M.D. Tenn. (song-lyric copyright)
    • Walters v. OpenAI (Gwinnett County Georgia 2023, dismissed May 2024 on First Amendment grounds where reasonable reader would not interpret ChatGPT output as factual)
    • Air Canada v. Moffatt BCCRT 2024 held the airline liable for representations made by its chatbot regarding bereavement-fare refunds, rejecting the argument that the chatbot was a “separate legal entity”
  • Centres on the duty of competence and candour to the tribunal.
    • Mata v. Avianca S.D.N.Y. (Judge Castel sanctions order, 22 June 2023) imposed $5,000 in Rule 11 sanctions against two attorneys who submitted ChatGPT-generated fake case citations including the fictitious “Varghese v. China Southern Airlines”
    • The case prompted federal-court standing orders requiring AI-use disclosure (Judge Starr N.D. Tex., Judge Pearson E.D. Ind., and over 40 federal judges by mid-2025)
    • ABA Formal Opinion 512 (July 2024) on generative AI in legal practice sets the duty-of-competence baseline
    • UK parallels include the Bar Standards Board guidance on AI use (December 2023) and the SRA’s AI Risk Outlook (2024)
    • The Harber v. HMRC First-tier Tribunal decision (December 2023) and Felicity Harber AI-citation case raised similar issues in UK tax tribunals
    • The Bar Council issued a Code of Conduct supplement on AI use (September 2024)

Discrimination and Algorithmic Bias

  • Encompasses claims under multiple regimes:
    • EU AI Act prohibited-practices regime (Article 5): social scoring, real-time biometric identification in public spaces with narrow exceptions, emotion recognition in workplaces and education, biometric categorisation by protected characteristics, predictive policing on profiling alone
    • Information Commissioner’s Office (ICO) enforcement: the Clearview AI fine of £7.5M (May 2022, reduced on appeal October 2023 on jurisdictional grounds), the Snapchat My AI investigation (2023, closed with assurances), the ongoing facial-recognition investigation into South Wales Police
    • FTC Section 5 enforcement: the December 2023 Rite Aid five-year ban on facial-recognition use after wrongful identification of shoppers as shoplifters disproportionately affecting Black and Latino customers
    • European Court of Human Rights jurisprudence on algorithmic decision-making in welfare administration (Glukhin v. Russia 2023) and biometric surveillance (Hambardzumyan v. Armenia 2024)
    • Loomis v. Wisconsin 881 N.W.2d 749 (Wis. 2016) on COMPAS risk-assessment use in sentencing

Consumer Protection and Deceptive AI Practices

  • The frontier of FTC/CMA enforcement.
    • The FTC’s Operation AI Comply (September 2024) brought actions against five companies for deceptive AI claims (DoNotPay’s “robot lawyer” claims, Ascend Ecom’s automated dropshipping claims, Ecommerce Empire Builders, FBA Machine, Rytr’s defamation-generating outputs)
    • The CFPB has issued circulars on AI in credit underwriting and adverse-action notices (Circular 2023-03 requiring specific reasons even when AI is used)
    • The CMA in the UK has examined foundation-model markets (CMA Foundation Models Initial Report September 2023, updated April 2024)
    • The European Commission opened formal investigations into Microsoft-OpenAI, Amazon-Anthropic and Google-Anthropic partnerships under the EU Merger Regulation
    • State AG enforcement: California, New York and Massachusetts have brought claims under state UDAP statutes for misrepresentations about AI capabilities

Academic Context

  • The academic study of AI liability draws on legal scholarship, computer science, economics, philosophy and public policy.
  • Ryan Calo, “Robotics and the Lessons of Cyberlaw” (California Law Review 103, 2015) framing the categorisation problem
  • Jack Balkin, “The Three Laws of Robotics in the Age of Big Data” (Ohio State Law Journal 78, 2017) on platform responsibility
  • Frank Pasquale, “The Black Box Society” (Harvard UP 2015) on algorithmic opacity
  • Margot Kaminski, “Binary Governance: Lessons from the GDPR’s Approach to Algorithmic Accountability” (Southern California Law Review 92, 2019)
  • Restatement (Third) of Torts: Products Liability by James Henderson and Aaron Twerski (American Law Institute 1998)
  • The in-progress Restatement of the Law, Torts: Concluding Provisions of the American Law Institute (drafts 2023-2025)

Frontier-AI Liability Theory

  • Gabriel Weil (Touro Law Center), “Tort Law as a Tool for Mitigating Catastrophic Risk from Artificial Intelligence” (Vermont Law Review 48, 2024)
  • Weil, “Your AI Breaks It? You Buy It” (Noema Magazine 2024) — accessible policy-focused presentation of the abnormally-dangerous-activity thesis
  • RAND Corporation, “U.S. Tort Liability for Large-Scale Artificial Intelligence Damages: A Primer for Developers and Policymakers” (RR-A3084-1, 2024) by Ketan Ramakrishnan
  • Anthony Aguirre, Gaia Dempsey et al., “Reinsuring AI: Energy, Agriculture, Finance & Medicine as Precedents for Scalable Governance of Frontier Artificial Intelligence” (FLI 2024)
  • Cullen O’Keefe (Institute for Law & AI), “Law-Following AI” project on building AI agents that comply with legal rules

Economic Theory of Liability

  • Polinsky & Shavell, “Punitive Damages: An Economic Analysis” (Harvard Law Review 111, 1998)
  • Shavell, “Foundations of Economic Analysis of Law” (Belknap 2004)
  • Anton Korinek & Joseph Stiglitz, “Artificial Intelligence and Its Implications for Income Distribution and Unemployment” (NBER WP 24174, 2017)
  • Tom Baker, “On the Genealogy of Moral Hazard” (Texas Law Review 75, 1996)
  • Kenneth Abraham, “Distributing Risk: Insurance, Legal Theory, and Public Policy” (Yale UP 1986)

UK and European Academic Centres

  • Cambridge Centre for Law & Future of Innovation — Felix Steffek on AI in dispute resolution, John Bell and Sarah Worthington on commercial law and AI agents
  • UCL Centre for Artificial Intelligence (joint Law/Computer Science) — Mireille Hildebrandt on legal protection by design, programme director Daniel Hulme on industry interface
  • Sandra Wachter (Oxford Internet Institute) — influential papers on counterfactual explanations, the right to reasonable inferences, and bias auditing
  • Edinburgh Centre for Technomoral Futures (Shannon Vallor) and the Bayes Centre — UK work on AI ethics
  • Imperial College Centre for Technology Foresight — systems-engineering perspectives on AI safety
  • Oxford Internet Institute (Luciano Floridi until 2023; Sandra Wachter, Brent Mittelstadt) and the now-closed Future of Humanity Institute (closed April 2024) drove the early frontier-AI risk literature
  • Alan Turing Institute Public Policy Programme — applied AI-governance research with UK regulators
  • Cambridge Centre for the Study of Existential Risk (CSER) and the Leverhulme Centre for the Future of Intelligence (CFI)

US Academic Centres

  • Stanford CodeX (Roland Vogl, Dan Ho) on legal informatics
  • Berkman Klein Center at Harvard (Yochai Benkler, Jonathan Zittrain) on platform governance
  • NYU AI Now Institute (Meredith Whittaker, Kate Crawford until 2022) on critical AI studies
  • Princeton Center for Information Technology Policy (Edward Felten, Arvind Narayanan)
  • Berkeley Law’s Technology, Society & Policy Initiative
  • Touro University Law Center — focal point for tort-theoretic approaches via Gabriel Weil
  • Institute for Law & AI (Cullen O’Keefe, formerly of OpenAI)
  • GovAI (Allan Dafoe, formerly Oxford, now Google DeepMind)
  • Yale Information Society Project (Jack Balkin)

Journals and Conferences

  • Journals: Yale Journal of Law & Technology, Harvard Journal of Law & Technology, Cambridge Law Journal, Modern Law Review, Oxford Journal of Legal Studies, Computer Law & Security Review, International Journal of Law and Information Technology, AI and Ethics (Springer), Journal of AI Law (launched 2024)
  • Conferences: FAccT (ACM Conference on Fairness, Accountability, and Transparency), AIES (AAAI/ACM AI Ethics & Society), the Society for Computers & Law annual conference, We Robot, ICAIL (International Conference on Artificial Intelligence and Law)

Current Landscape (2026)

EU AI Act Implementation

  • Regulation (EU) 2024/1689 published 12 July 2024 entered into force 1 August 2024
  • Prohibitions on eight categories of prohibited AI practices applied from 2 February 2025: social scoring, manipulative subliminal techniques, exploitation of vulnerabilities, untargeted facial-image scraping, emotion recognition in workplaces and education, biometric categorisation by protected characteristics, real-time biometric identification in public spaces, predictive policing on profiling alone
  • GPAI obligations commenced 2 August 2025
  • The EU GPAI Code of Practice published May 2025 covers transparency (technical documentation, training-data summary), copyright (machine-readable opt-out compliance with Article 4 of Directive (EU) 2019/790), and safety/security (systemic-risk evaluation, incident reporting)
  • High-risk system obligations apply 2 August 2026
  • Penalties: €35M or 7% of global turnover for prohibited practices; €15M or 3% for high-risk non-compliance; €7.5M or 1% for incorrect information
  • The European AI Office (within DG CONNECT, established February 2024) co-ordinates GPAI oversight with the European Artificial Intelligence Board
  • Member-State National Competent Authorities designated by 2 August 2025

EU Revised Product Liability Directive

  • Directive (EU) 2024/2853 entered into force 9 December 2024
  • Member-State transposition required by 9 December 2026
  • Key changes from Directive 85/374/EEC:
    • “Product” expressly includes software, AI systems, digital-manufacturing files and “related services” integrated into a product
    • The €500 deductible is abolished
    • Damages now include destruction or corruption of non-professional data and medically recognised psychological harm
    • Article 10 introduces presumptions of defectiveness and causation when claimants face excessive evidentiary difficulties
    • Article 16 establishes disclosure obligations on defendants
    • The 10-year long-stop is extended to 25 years for latent-injury claims
    • The development-risks/state-of-the-art defence is retained but narrowed

AILD Withdrawal

  • The European Commission withdrew the AI Liability Directive proposal COM(2022) 496 on 11 February 2025
  • Withdrawal occurred as part of the 2025 Work Programme reset, citing “no foreseeable agreement”
  • The withdrawal leaves AI-specific causation rules to Member-State tort regimes
  • The revised PLD’s parallel presumptions cover the same evidentiary territory for product-liability claims
  • Academic commentary (Oxford Business Law Blog April 2025) argues EU law remains influential through the PLD even without the AILD
  • The Commission’s announcement signalled possible revisitation if implementation experience reveals gaps

US Federal Landscape

  • No comprehensive federal AI liability statute as of mid-2026
  • The Trump Administration’s “America’s AI Action Plan” (released July 2025) emphasises competition with China, seeks to pre-empt state regulations that “unduly restrict innovation,” and accelerates federal AI adoption
  • The bipartisan Senate AI Working Group’s May 2024 Roadmap remains a reference document
  • Federal agencies act under existing authorities: FTC Section 5, CFPB Dodd-Frank powers, EEOC Title VII, HUD Fair Housing Act, FDA FDC Act, NHTSA Motor Vehicle Safety Act
  • NIST published the AI RMF Generative AI Profile (NIST AI 600-1) in July 2024
  • The Executive Order on AI (EO 14110, October 2023) was rescinded January 2025 and replaced by the Trump AI Action Plan
  • The US AI Safety Institute (NIST AISI) survived the reorganisation in modified form

US State Landscape

  • Colorado AI Act SB 24-205 (enacted May 2024, effective February 2026) — first comprehensive state AI law covering consequential decisions in employment, housing, credit, healthcare, education, government services, legal services and insurance
  • Utah AI Policy Act (effective May 2024) requires disclosure for regulated occupations
  • California SB 1047 (passed Legislature August 2024) vetoed by Governor Newsom 29 September 2024; SB 53 reintroduced 2025
  • Texas Responsible AI Governance Act HB 1709 advanced through 2025
  • Connecticut SB 2 enacted 2024
  • Illinois HB 3773 effective January 2026 covers AI in employment
  • New York City Local Law 144 continues in force; New York State S7623-A pending
  • Rhode Island S0358 (2025) introduces frontier-AI strict liability — Gabriel Weil instrumental in drafting

UK Trajectory

  • The pro-innovation framework continues but is contested
  • The Labour Government elected July 2024 committed in its manifesto to “binding regulation on the handful of companies developing the most powerful AI models”
  • No Government AI Bill had been introduced by mid-2026
  • Lord Holmes’ Artificial Intelligence (Regulation) Bill [HL] was reintroduced March 2025
  • The Department for Science, Innovation and Technology (DSIT) Central Function continues co-ordinating regulator action
  • The UK AI Safety Institute was renamed the AI Security Institute in February 2025 with narrower remit focused on national-security-related risks
  • The Automated Vehicles Act 2024 (Royal Assent 20 May 2024) is the most significant AI-specific UK statute, implementing the joint Law Commission/Scottish Law Commission proposals on the ASDE framework
  • The Online Safety Act 2023 implementation by Ofcom contains AI-relevant provisions on illegal content and child-safety duties

Notable 2024-2026 Cases

  • Air Canada v. Moffatt 2024 BCCRT 149 (chatbot misrepresentation; airline liable)
  • Mata v. Avianca SDNY 22 June 2023 (Rule 11 sanctions of $5,000 for ChatGPT-fabricated citations)
  • Garcia v. Character Technologies M.D. Fla. 21 May 2025 (chatbot app as product for strict liability)
  • Getty Images v. Stability AI UK High Court trial commenced June 2025
  • Andersen v. Stability AI N.D. Cal. motion-to-dismiss ruling 12 August 2024
  • New York Times v. OpenAI S.D.N.Y. filed December 2023
  • Walters v. OpenAI Gwinnett County 2024 (defamation by ChatGPT, dismissed 19 May 2024 on First Amendment grounds)
  • In re Social Media Adolescent Addiction MDL 3047 N.D. Cal. (consolidated cases against Meta, TikTok, Snap, YouTube on platform-design product liability)
  • Mobley v. Workday N.D. Cal. (Workday recruitment AI alleged discrimination)
  • Concord Music v. Anthropic M.D. Tenn. (song-lyric copyright)

FTC and Federal Enforcement

  • Rite Aid five-year facial-recognition ban (December 2023) remains the high-water mark — first FTC ban on a specific AI use
  • Operation AI Comply (September 2024) targeted five firms for deceptive AI claims
  • Amazon Alexa COPPA settlement ($25M, 2023) addressed children’s voice-recording retention
  • Ring settlement ($5.8M, 2023) addressed employee access to consumer video
  • CFPB Circular 2023-03 requires specific reasons in adverse-action notices even when AI is used
  • NIST AI Safety Institute Consortium (AISIC) formed May 2024 with 200+ member organisations

Insurance Market

  • Premium volume estimated at 4-6B by 2030
  • Munich Re aiSure expanded for generative AI 2024
  • Lloyd’s of London syndicates Tokio Marine Kiln, Beazley, Chaucer, Hiscox write affirmative AI coverage
  • InsurTech specialists Vouch and Armilla AI offer specialist policies
  • Cyber carriers Coalition and At-Bay introduced cyber+AI endorsements
  • AIG’s AI risk addendum to professional indemnity launched 2024
  • Swiss Re, Hannover Re, SCOR provide reinsurance backing
  • Lloyd’s-led market forecasts suggest AI catastrophe coverage gap requiring government reinsurance backstop by 2030

UK Context

  • The United Kingdom’s approach to AI liability blends a deliberately light-touch principles-based regulatory framework with statutory action in specific sectors and active academic-legal scholarship.

Pro-Innovation Regulatory Framework

  • The March 2023 White Paper “A pro-innovation approach to AI regulation” (CP 815) declined to legislate a horizontal AI statute
  • Existing regulators are tasked with applying five cross-cutting principles within their existing remits:
    • Safety, security and robustness
    • Appropriate transparency and explainability
    • Fairness
    • Accountability and governance
    • Contestability and redress
  • Key sectoral regulators applying the principles:
    • Information Commissioner’s Office (ICO) — data protection and AI
    • Competition and Markets Authority (CMA) — competition and consumer
    • MHRA — medical devices and AI
    • Financial Conduct Authority (FCA) — financial-services AI
    • Prudential Regulation Authority (PRA) — bank/insurer model risk
    • Ofcom — online safety and AI content
    • Health and Safety Executive (HSE) — workplace AI
    • Equality and Human Rights Commission (EHRC) — discrimination
    • Office for Standards in Education (Ofsted) — education AI
  • The Department for Science, Innovation and Technology (DSIT) operates a Central Function to co-ordinate cross-regulator activity

Parliamentary Pressure for Statutory Footing

  • The House of Commons Science, Innovation and Technology Committee has challenged the light-touch approach
  • August 2023 interim report “The governance of artificial intelligence”
  • August 2024 Final Report called for statutory footing and clear allocation of liability via legislation
  • Lord Holmes of Richmond introduced the Artificial Intelligence (Regulation) Bill [HL] in the 2023-24 session (Bill 11)
  • The Bill was reintroduced as a Private Member’s Bill in March 2025
  • Provisions include: establishing an AI Authority, mandating regulatory sandboxes, requiring AI-impact assessments
  • The Labour Government elected July 2024 committed to legislating for “the most powerful AI models” but no Government Bill had been introduced as of mid-2026

Automated Vehicles Act 2024

  • Royal Assent 20 May 2024 — the most significant AI-specific UK statute
  • Implements the joint Law Commission of England and Wales / Scottish Law Commission Automated Vehicles project (final joint report January 2022, HC 1149)
  • Creates the Authorised Self-Driving Entity (ASDE) framework
  • An ASDE bears criminal and civil responsibility for the driving behaviour of a self-driving vehicle while engaged in self-driving mode
  • Displaces the user-in-charge’s liability during self-driving operation
  • Statutory Instruments under the Act will implement the In-Use Regulatory Scheme administered jointly by DfT, DVSA and the new regulator
  • Phased implementation 2025-2027

ICO as Active AI Regulator

  • The most active UK AI regulator with broadest enforcement reach
  • Key guidance: Guidance on AI and Data Protection (updated 2023), the AI auditing framework (2020), and the Generative AI consultation series (2024)
  • Enforcement: the Clearview AI £7.5M fine (May 2022, reduced on appeal October 2023 on jurisdictional grounds)
  • Snapchat My AI investigation (2023, closed with assurances)
  • Ongoing investigation into facial-recognition use by South Wales Police
  • The ICO operates the AI and Biometrics Strategy with the EHRC

CMA Foundation Models Work

  • CMA Initial Report on AI Foundation Models (September 2023)
  • Updated Principles (April 2024): Accountability, Access, Diversity, Choice, Flexibility, Fair Dealing, Transparency
  • Merger reviews: Microsoft-OpenAI (December 2023, closed March 2024 finding no relevant merger situation), Amazon-Anthropic (closed September 2024), Google-Anthropic (closed November 2024)
  • The CMA’s market study powers under the Enterprise Act 2002 enable continuing oversight

MHRA Software and AI as Medical Device

  • MHRA Software and AI as Medical Device Change Programme announced 2021
  • Roadmap published 2024 covering 11 work-packages
  • Topics include qualification, classification, premarket requirements, post-market surveillance, cybersecurity and AI-specific provisions
  • AI Airlock regulatory sandbox launched May 2024 for adaptive AI medical devices
  • Alignment with IMDRF guidance and IEC 62304/82304

UK AI Safety Institute → AI Security Institute

  • Established as part of the November 2023 Bletchley Park AI Safety Summit
  • Conducted pre-deployment evaluations of frontier models with voluntary cooperation from OpenAI, Anthropic, Google DeepMind, Meta, Mistral, Microsoft and Amazon
  • Renamed the AI Security Institute (AISI) in February 2025 with narrowed remit
  • New focus: national-security-related risks (cyber, CBRN, autonomous-systems abuse)
  • The AISI’s evaluations framework continues to influence international model-testing standards
  • International Network of AI Safety Institutes coordinated via the Seoul Declaration (May 2024) and Paris Action Summit (February 2025)

UK Academic Centres

  • Cambridge Centre for Law & Future of Innovation (LawTech Lab, Felix Steffek) on AI in dispute resolution, AI contracting and digital assets
  • UCL Centre for Artificial Intelligence (Computer Science + Laws) houses Mireille Hildebrandt’s work on legal protection by design
  • Edinburgh Centre for Technomoral Futures (Shannon Vallor) leads UK work on AI ethics
  • Imperial College Centre for Technology Foresight contributes systems engineering perspectives
  • Oxford Internet Institute (Sandra Wachter, Brent Mittelstadt) — counterfactual-explanations and right-to-reasonable-inferences scholarship
  • Alan Turing Institute Public Policy Programme — applied research with UK regulators

Northern English Industrial AI-Law Cluster

  • Manchester’s School of Law — David Sugarman, Diane Coyle on AI economics
  • Leeds’ Information School — Helen Kennedy on algorithmic accountability
  • Sheffield’s Information School and Newcastle’s School of Law contribute the Northern English industrial AI-law cluster
  • Manchester serves as the AI Greater North hub with industrial AI deployments in:
    • Transport: TfGM, Northern Trains operational AI
    • Healthcare: Manchester University NHS Foundation Trust AI pathology
    • Advanced manufacturing: Sheffield AMRC AI for additive manufacturing
    • Law: Manchester legaltech cluster including Slaughter and May’s Luminance investment

Law Commission and Professional Bodies

  • The Law Commission of England and Wales continues scoping work on AI liability, building on the Automated Vehicles project
  • The Commission’s 14th Programme of Law Reform (March 2024) includes provisional projects on Digital Assets and ETDs
  • Bar Standards Board AI use guidance (December 2023)
  • SRA Risk Outlook (2024) on AI in solicitors’ practice
  • The Society for Computers & Law convenes the leading UK AI-law practitioner community
  • The British Computer Society (BCS) and Royal Statistical Society contribute professional-standards work

Future Directions (2026-2030)

  • Five trajectories will define AI liability through 2030. First, horizontal AI liability legislation is likely in the UK before end-2027 covering frontier-model developers with mandatory pre-deployment evaluation, model registration, and post-deployment incident reporting; the Government’s stated commitment to “binding regulation on the handful of companies developing the most powerful AI models” plus mounting Parliamentary pressure make a Government AI Bill increasingly probable. Second, the EU AI Act high-risk regime comes fully into force 2 August 2026 producing the first wave of conformity-assessment litigation and Member-State market-surveillance enforcement, with national competent authorities designated by 2 August 2025 and the European AI Office (within DG CONNECT, established within the European Commission February 2024) co-ordinating GPAI oversight. Third, AI liability litigation in the US will expand dramatically as MDL 3047 (social-media adolescent addiction) and the foundation-model copyright cases (NYT v. OpenAI, Andersen v. Stability AI, Authors Guild v. OpenAI) reach summary-judgment and trial stages 2026-2028; appellate review of Character.AI’s product-liability ruling will set a critical precedent on whether chatbot apps are “products” for strict-liability purposes. Fourth, insurance market formalisation will accelerate, with Lloyd’s expected to publish AI-specific model wordings, the IUA producing standard exclusions/inclusions, and reinsurers (Munich Re, Swiss Re, Hannover Re, SCOR, Lloyd’s) developing AI-catastrophe coverage; mandatory professional indemnity for AI deployers in regulated sectors (healthcare, finance, legal) is probable by 2028. Fifth, frontier-AI strict liability debates will intensify as larger models deploy: the Rhode Island S0358 bill or equivalent state legislation may pass; California SB 53 (the SB 1047 successor) faces revived consideration; UK Parliamentary debate on abnormally-dangerous-activity treatment for frontier model development will gather force; and the EU may revisit the AILD withdrawal in light of accumulated PLD case law.
  • Beyond these near-term trajectories, three deeper structural shifts are emerging. The convergence of AI liability and AI safety: pre-deployment evaluation requirements (initially voluntary under the Bletchley/Seoul/Paris commitments and the EU GPAI Code of Practice) will harden into legally binding obligations, creating an audit-trail prerequisite for limitation of liability defences. The rise of compulsory insurance regimes modelled on the Price-Anderson Act’s nuclear-industry precedent, with industry pools and government reinsurance backstops; the OMS Insurance “Insuring emerging risks from AI” framework (November 2024) and the ResearchGate Reinsuring AI proposals (Aguirre et al. 2024) sketch the design. The judicialisation of AI safety: as case law accumulates, technical standards (ISO/IEC 42001, NIST AI RMF, the EU GPAI Code of Practice) will become de facto standards of care, with breach producing presumptive negligence under the negligence per se doctrine; conversely, certification under recognised frameworks will become a partial safe-harbour defence.
  • The deepest unresolved question is whether existing tort doctrines can be stretched to cover frontier AI, or whether purpose-built regimes are required. The dominant academic view (Weil, Korinek, Aguirre, the RAND analysis) is that catastrophic and existential risks from frontier AI cannot be adequately deterred by compensatory damages alone—an insurance-uninsurable, judgment-proof scenario where no plaintiff exists post-catastrophe. Solutions include Weil’s warning-shot punitive multipliers (calculating punitive damages on near-miss events to reflect the probability-weighted cost of the unrealised catastrophic risk), mandatory compulsory insurance (forcing insurers to become quasi-regulators of safety practices), and strict liability for abnormally dangerous activity treating frontier-model training as ultrahazardous. Whether any of these is politically achievable—particularly in the US where federal AI legislation faces structural obstacles and the Trump Administration prioritises deregulation—remains the defining open question of AI governance.

Research & Literature

  • Core academic works. Gabriel Weil, “Tort Law as a Tool for Mitigating Catastrophic Risk from Artificial Intelligence” (Vermont Law Review 48, 2024); Ketan Ramakrishnan et al., “U.S. Tort Liability for Large-Scale Artificial Intelligence Damages” (RAND RR-A3084-1, 2024); Margot Kaminski, “Binary Governance: Lessons from the GDPR’s Approach to Algorithmic Accountability” (Southern California Law Review 92, 2019); Mireille Hildebrandt, “Law for Computer Scientists and Other Folk” (OUP 2020); Sandra Wachter & Brent Mittelstadt, “A Right to Reasonable Inferences” (Columbia Business Law Review 2019); Ryan Calo, “Robotics and the Lessons of Cyberlaw” (California Law Review 103, 2015); Frank Pasquale, “The Black Box Society” (Harvard UP 2015); Jack Balkin, “The Three Laws of Robotics in the Age of Big Data” (Ohio State Law Journal 78, 2017); Anton Korinek & Joseph Stiglitz, “Artificial Intelligence and Income Distribution” (NBER 24174, 2018).
  • Key reports. RAND Corporation “U.S. Tort Liability for Large-Scale AI Damages” RR-A3084-1 (2024); House of Commons Science, Innovation and Technology Committee “The governance of artificial intelligence” Final Report (August 2024); UK Government “A pro-innovation approach to AI regulation” White Paper CP 815 (March 2023) and Consultation Response (February 2024); European Commission “AI Liability Directive Withdrawal Notice” (11 February 2025); RAND “Insuring Emerging Risks from AI” OMS Insurance (November 2024); Aguirre, Dempsey et al., “Reinsuring AI” Future of Life Institute (2024); CMA “AI Foundation Models: Initial Report” (September 2023) and “AI Foundation Models: Update Paper” (April 2024).
  • Primary legislation. Regulation (EU) 2024/1689 (EU AI Act); Directive (EU) 2024/2853 (revised Product Liability Directive); UK Automated Vehicles Act 2024; UK Consumer Protection Act 1987; US Restatement (Second) of Torts § 402A; US Restatement (Third) of Torts: Products Liability (1998); Colorado AI Act SB 24-205; New York City Local Law 144 of 2021; Rhode Island Senate Bill S0358 (2025).
  • Landmark cases. Donoghue v. Stevenson [1932] AC 562; Caparo Industries plc v. Dickman [1990] 2 AC 605; Rylands v. Fletcher (1868) LR 3 HL 330; Brown v. Kendall 60 Mass 292 (1850); Palsgraf v. Long Island Railroad 248 NY 339 (1928); BMW v. Gore 517 US 559 (1996); State Farm v. Campbell 538 US 408 (2003); Rookes v. Barnard [1964] AC 1129; Daubert v. Merrell Dow Pharmaceuticals 509 US 579 (1993); Mata v. Avianca No. 22-CV-1461 (SDNY 22 June 2023); Air Canada v. Moffatt 2024 BCCRT 149; Garcia v. Character Technologies Inc. M.D. Fla. 21 May 2025; Andersen v. Stability AI No. 3:23-cv-00201 (N.D. Cal. 12 August 2024); Getty Images v. Stability AI UK High Court [2025] EWHC pending; Walters v. OpenAI No. 23-A-04860-2 (Gwinnett County GA dismissed 19 May 2024).
  • Technical standards. ISO/IEC 42001:2023 (AI Management System); ISO/IEC 23894:2023 (AI risk management); ISO/IEC TR 24028 (AI trustworthiness); IEC 62304/82304 (medical software lifecycle); NIST AI RMF 1.0 (January 2023) with Generative AI Profile NIST AI 600-1 (July 2024); IEEE 7000-2021 (Model Process for Addressing Ethical Concerns During System Design); IEEE 7001-2021 (Transparency of Autonomous Systems); IMDRF/SaMD WG (medical software).
  • Practitioner sources. Hogan Lovells “AI Liability in the EU and UK” (2024-2025); Norton Rose Fulbright “Artificial Intelligence and Liability” (2024); Morrison Foerster “Software Gains New Status as a Product Under Strict Liability Law” (2025); Bird & Bird “EU AI Act: Practical Guide” (2024); Linklaters “AI Liability Tracker” (quarterly 2024-2026); DLA Piper “Insuring the unpredictable: The challenges of AI risk insurability” (2024); American Bar Association Section of Litigation “Have Algorithms Opened Up Your Software to Product Liability?” (2024); Stanford Encyclopaedia of Philosophy entry on “Ethics of Artificial Intelligence and Robotics” (Müller, revised 2024).

Metadata

Provenance

Graph View

Backlinks