The AI Liability Directive was a 2022 European Commission proposal (COM/2022/596) that would have made it easier for people harmed by artificial intelligence to win compensation in civil court. The Commission withdrew it in October 2025 after EU legislative bodies could not agree on the text, and no replacement has been announced. For anyone harmed by an AI system in the EU today, the practical answer sits elsewhere: the revised Product Liability Directive now treats software and AI as products subject to strict liability, and national tort law covers what the PLD does not.
What the Proposal Would Have Changed
The directive targeted a specific evidentiary problem. Under ordinary fault-based liability, a person claiming harm must prove someone was negligent and that the negligence caused the injury. With AI, both steps are extraordinarily difficult. The system’s internal decision-making is opaque even to the engineers who built it, and tracing a specific output back through layers of algorithmic processing can be functionally impossible.
It would have applied to non-contractual civil claims, meaning situations where the harmed person has no direct contract with the AI provider or the business using the system. The rules covered both providers, who develop AI systems, and deployers, who use them professionally. Different evidentiary standards applied depending on whether the AI counted as “high-risk” under the companion EU AI Act, a category that includes systems used in recruitment, credit scoring, law enforcement, biometric identification, critical infrastructure, and the administration of justice.1EU AI Act. EU AI Act – Annex III
The Rebuttable Presumption of Causality
The most important feature was a rebuttable presumption of causality. If certain conditions were met, a court could assume the AI system caused the harm instead of forcing the victim to reverse-engineer the algorithm to prove it.
The presumption was not automatic. A claimant would have needed to show three things: that the defendant failed to comply with a relevant duty of care such as a safety or transparency requirement; that this failure reasonably could have influenced the AI’s output; and that the output was connected to the actual damage.2European Parliament. Artificial Intelligence Liability Directive For high-risk AI, the presumption would have triggered more readily, particularly where a provider ignored AI Act requirements on data governance, record-keeping, or human oversight. For non-high-risk systems, courts would apply the presumption only when proving causation was excessively difficult for the victim. Defendants could rebut the presumption with evidence pointing to a different cause. This was not strict liability; the claimant still had to establish negligence.
Evidence Disclosure
The proposal also gave national courts the power to order providers and deployers of high-risk AI systems to disclose internal records, logs, and technical specifications, once the claimant had presented enough facts to make the claim plausible. Disclosure was limited to what was “necessary and proportionate,” with trade secret safeguards built in. If a defendant refused to comply, a presumption of non-compliance kicked in, meaning the court could assume the withheld evidence would have shown a duty-of-care violation. Requests were expected to focus on risk management systems and the training datasets used to build the model.
Why the Directive Was Withdrawn
The Commission published notice of the withdrawal in the Official Journal on October 6, 2025. The stated reason was the lack of a foreseeable agreement among the institutions. The Commission said it would consider whether to present a revised proposal or take a different approach; as of early 2026, none has been announced.
The failure was not a rejection of the underlying diagnosis. The European Parliament’s own analysis acknowledged that AI’s opacity, autonomous behavior, and continuous adaptation make it “particularly difficult” for victims to meet the burden of proof under existing rules.2European Parliament. Artificial Intelligence Liability Directive Legislators disagreed on the details, not the problem. So the problem remains, without harmonized EU-wide rules to address it.
What Law Now Applies to AI Harm in the EU
The revised Product Liability Directive (Directive (EU) 2024/2853), which member states must transpose into national law by December 9, 2026, explicitly treats software and AI systems as “products” subject to strict liability.3EUR-Lex. Directive (EU) 2024/2853 – Product Liability Directive This is a different legal tool than the withdrawn proposal offered, and in some ways a stronger one.
Under strict liability, the harmed person does not need to prove the manufacturer was negligent. They need to show the product was defective, that they suffered damage, and that the defect caused the damage. A product is defective when it does not provide the safety a person is entitled to expect.3EUR-Lex. Directive (EU) 2024/2853 – Product Liability Directive
The revised PLD carries its own presumptions to help claimants. Defectiveness is presumed when a manufacturer fails to disclose required information, when the product violates mandatory safety requirements, or when the damage was caused by an obvious malfunction. A causal link between defect and damage is presumed when the damage is typically consistent with the type of defect in question, or when technical complexity makes proving the link excessively difficult.4European Parliament. Revised Product Liability Directive That last provision directly addresses the black-box problem the withdrawn directive was meant to solve.
Disclosure obligations are in the PLD too. Once a claimant presents enough facts to support the plausibility of a claim, the manufacturer must produce necessary information, subject to trade secret protections.4European Parliament. Revised Product Liability Directive Software developers, including AI system providers as defined under the EU AI Act, count as manufacturers for liability purposes.3EUR-Lex. Directive (EU) 2024/2853 – Product Liability Directive
What the PLD Does Not Cover
The PLD reaches personal injury, property damage, and data loss. It does not reach infringements of personality rights or pure financial loss. If an AI hiring tool discriminates against you and you lose a job opportunity but suffer no physical harm or property damage, the PLD may not help. It also includes a state-of-the-art defense: manufacturers can avoid liability for defects that could not have been discovered through the scientific and technical knowledge available when the product entered the market. Free and open-source software developed outside a commercial activity is excluded, though that exception disappears once someone integrates the code into a commercial product.
Where the EU AI Act Fits
The EU AI Act (Regulation (EU) 2024/1689) is a separate framework. It sets safety requirements to reduce risk before harm occurs, and contains no provisions creating a right to compensation after harm happens.2European Parliament. Artificial Intelligence Liability Directive
It matters for liability anyway. Providers of high-risk AI systems must implement risk management systems, use high-quality training datasets, maintain technical documentation, enable automatic event logging, provide instructions for deployers, design for human oversight, and meet standards for accuracy, robustness, and cybersecurity.5Artificial Intelligence Act. High-Level Summary of the AI Act When a provider fails one of these obligations, that failure becomes powerful evidence in a product liability or tort claim under other laws.
Reach Beyond Europe
Neither the AI Act nor the revised PLD is limited to European companies. The AI Act applies to providers anywhere in the world if they place AI systems on the EU market or put them into service in the EU, and to both providers and deployers when the output of an AI system is used in the EU.6Morgan Lewis. The EU Artificial Intelligence Act Is Here – With Extraterritorial Reach
The PLD reaches non-EU manufacturers through a chain of liability. Where a manufacturer is established outside the EU, liability falls on the importer, the manufacturer’s authorized EU representative, or, failing those, the fulfillment service provider.3EUR-Lex. Directive (EU) 2024/2853 – Product Liability Directive A U.S. company selling AI software to European customers cannot avoid exposure by lacking a European headquarters; someone in the supply chain will bear the claim.
The Gaps the Withdrawal Leaves
The revised PLD is a strong tool for defective products, but it does not cover every AI harm. Discriminatory outcomes in hiring or lending, privacy violations by automated systems, and purely financial losses from algorithmic errors fall outside its scope. For those harms, victims must rely on national tort law, which varies significantly across the EU. Some member states apply strict liability to certain dangerous activities; others use fault-based frameworks where the victim carries the full burden of proving negligence and causation.
That fragmentation is exactly what the AI Liability Directive was meant to fix. Without harmonized rules, a person harmed by the same AI system in the same way may have a viable claim in one member state and no realistic path to compensation in another. The Commission has said it may return with a revised proposal or an alternative approach, but no timeline exists. For now, anyone affected by AI-related harm in the EU should look first to the revised Product Liability Directive where it applies, and to national tort law where it does not.