EU AI Liability Directive (Proposed)
The proposed EU AI Liability Directive would ease the burden of proof for victims of AI-caused harm through evidence-disclosure rights and a rebuttable presumption of causality. Its status is uncertain after the Commission signaled a possible withdrawal.
Overview
The EU AI Liability Directive was proposed by the European Commission in September 2022 as a companion to the EU AI Act. Where the AI Act sets safety and governance rules, the AI Liability Directive addresses what happens when AI systems cause harm. Its goal is to make it easier for individuals and businesses to claim compensation for damage caused by AI, given that the opacity and autonomy of AI systems can make traditional fault-based claims very hard to prove.
The proposal harmonizes certain national civil-liability rules for non-contractual, fault-based claims involving AI. It is important to track this measure carefully: the Commission signaled in 2025 that it intended to withdraw the proposal, and its future is uncertain. It is included here as a real and influential regulatory instrument whose status is in flux.
Who It Applies To
The directive, if adopted, would apply to claims for damage caused by AI systems within the EU. It would benefit any natural or legal person who suffers harm and would impose evidentiary and disclosure obligations on providers and users of AI systems, including businesses deploying high-risk AI as defined by the AI Act.
Key Requirements
Two mechanisms are central. First, a disclosure right would let courts order providers of high-risk AI systems to disclose relevant evidence, such as logs and technical documentation, to claimants who have presented a plausible claim. Second, a rebuttable presumption of causality would ease the burden of proving the causal link between a defendant's fault and the AI output that caused harm, where the claimant shows non-compliance with a relevant duty of care.
Penalties for Non-Compliance
The directive is not a penalty regime. It governs civil liability and compensation between private parties. Its practical effect is to increase litigation exposure for AI providers and deployers by lowering the evidentiary hurdles claimants face. Failure to preserve or disclose required evidence can trigger an adverse presumption against the defendant.
How to Comply
Organizations should maintain thorough technical documentation, logging, and traceability for AI systems, especially high-risk ones, so they can respond to disclosure orders and rebut causality presumptions. Aligning AI governance with the EU AI Act and maintaining strong records of testing, monitoring, and risk management is the most effective preparation, regardless of the directive's final fate.