Skip to main content
NormaGrid

AI Liability Directive (AILD)

In process AI Proposed Directive
This proposal has been withdrawn and is no longer being pursued. This page is kept for reference.

AI-assisted content notice: this page includes AI-assisted summaries, FAQs, and glossary entries prepared for navigation purposes. Verify the underlying legal text before relying on this content.

Current position & sources

Sources checked:

The AI Liability Directive proposal COM(2022)496 was withdrawn before adoption. The Commission approved withdrawal on 16 July 2025 and published the notice on 6 October 2025. Its proposed evidence and causality rules did not become binding law.

Summary

The AI Liability Directive proposal COM(2022)496 was withdrawn before adoption. The Commission approved withdrawal on 16 July 2025 and published the notice on 6 October 2025. Its proposed evidence and causality rules did not become binding law.

Who is affected?

The proposal concerned people seeking compensation for AI-related harm, potentially liable providers and users, and national courts. It creates no current obligations for those groups.

Scope

Historical proposal for selected procedural rules in non-contractual, fault-based civil claims involving AI systems. It was not a general strict-liability regime.

Key Points

  • The Commission document is dated 28 September 2022; Parliament records publication on 29 September 2022.
  • Courts would have been able to order targeted disclosure or preservation of evidence concerning specific high-risk AI systems.
  • A rebuttable causality presumption would have eased certain evidentiary burdens, subject to conditions.
  • Withdrawal ended this proposal without an adoption, application or transposition deadline.
  • Withdrawal does not remove liability under other applicable EU or national law.

Key Deadlines

  • — Commission proposal document dated
  • — Proposal publication recorded by Parliament
  • — Commission approved withdrawal
  • — Withdrawal notice published in the Official Journal

Related Regulations

Frequently Asked Questions

Is this directive in force?

No. It was withdrawn before adoption and never created binding duties.

What did the proposal try to change?

It addressed evidence in certain AI-related civil claims through disclosure and rebuttable presumptions.

Did it make every AI provider strictly liable?

No. Its focus was fault-based non-contractual claims, with conditions on the proposed procedural tools.

When was it withdrawn?

The Commission approved withdrawal on 16 July 2025; the Official Journal notice was published on 6 October 2025.

Does withdrawal prevent claims for AI-related harm?

No. Applicable liability rules must still be assessed for the claim concerned.

Is there a compliance or transposition deadline?

No deadline arose from this withdrawn proposal. Dates on this page describe its legislative history.

Key Terms

Fault-based liability
Liability that depends on a legally relevant fault, rather than merely on the involvement of an AI system.
Non-contractual claim
A claim for harm that is not founded on a contractual obligation.
Disclosure of evidence
Court-ordered access to relevant evidence; the proposal envisaged a targeted and proportionate mechanism.
Rebuttable presumption
An evidentiary assumption that the opposing party may challenge.
Withdrawal
Termination of a legislative proposal before adoption; distinct from repeal of an enacted law.