Significantly higher requirements for restitution and damage compensation claims than previously assumed
On 28 September, the Commission published its proposal for an AI Liability Directive. We previously reported on a leaked version. The wording and content of the version now published differs from the previous version in some relevant points. We therefore present the main contents of the proposal here.
Scope of the AI Liability Directive
The AI Liability Directive applies only to non-contractual fault-based damage compensation claims. Any damage compensation claims resulting from the Product Liability Directive and the liability exemptions and due diligence obligations from the Digital Services Act (planned EU regulation for a law on digital services) are to remain unaffected. The essential definitions of terms from related legal acts of the European Union, in particular the AI Regulation, have been adopted.
Disclosure obligations of the operators and users of AI systems
In accordance with Article 3 of the draft AI Liability Directive, a potential claimant in the event of damage should first request the disclosure of relevant evidence from the operator of the AI system or persons equivalent to the operator. This requirement does not apply if the damage compensation claim is brought before a court. In this case, just as when the operator refuses to disclose information, the courts have the power to order disclosure. For this purpose, however, the plaintiff must present sufficient facts and evidence to make the damage compensation claim plausible. The order also requires that the claimant has done everything reasonable to obtain evidence from the respondent.
If the surrender is refused, the court is to assume a breach of the duty on the part of the claimant to exercise due diligence and thus its probative value for the damage compensation claim. This presumption is rebuttable.
Reversal of burden of proof
Under the following three conditions, a causal link between the defendant’s fault and the AI system is to be (rebuttably) presumed by the court according to Article 4 of the draft:
- The court presumes (due to non-disclosure) or the plaintiff proves that the defendant breached duties to exercise due diligence that were precisely intended to prevent the damage that occurred.
- There is a reasonable probability that the breach of the duties to exercise due diligence impacted the harmful effects of the AI system.
- The plaintiff proves that the harmful effects of the AI system caused the harm.
If these conditions are met, the defendant has the burden of proving that the defendant is not responsible for the damage. However, the shift in the burden of proof will not apply if the defendant demonstrates that the claimant has sufficient evidence and expertise to prove causation.
The numerous changes between the leaked version of the AI Liability Directive and the draft now published demonstrate the highly dynamic situation surrounding the legal regulation of AI. The official draft of the AI Liability Directive mitigates legal risks for companies using AI applications. In principle, however, risks such as a loss of intellectual property due to disclosure obligations or a need to furnish evidence as a result of a reversal of the burden of proof remain. It also cannot be ruled out that the draft will be tightened up again in discussions with the EU Parliament and the Council of the European Union. Companies that use AI or plan to do so in the future should therefore monitor further developments closely and take any measures necessary to reduce liability at an early stage.back