The draft Directive adapting the rules on non-contractual civil liability to artificial intelligence (COM/2022/496; hereinafter: "AI Liability Directive") intends to extend and modernise the legal framework in the EU.  For the first time, specific rules are introduced for damages caused by AI systems. This is intended to promote trustworthy AI systems, ensure benefits for the internal market and eliminate or reduce legal uncertainties related to liability risks for companies.

The new rules are intended to ensure that victims of harm caused by AI- systems are compensated in the same way as in the case of other harmful circumstances. Among other points, the directive introduces two essential measures: The presumption of causality relieves victims of the obligation to set out and prove in detail how harm was caused by a particular fault or omission. It also allows access to evidence held by companies or providers, especially in the context of high-risk AI.

• The introduction of the AI Liability Directive poses a number of challenges for companies, including:
• Reversal of the burden of proof: The AI Liability Directive provides for a reversal of the burden of proof under certain conditions to make it easier for injured parties to bring claims against companies. When the reversal of the burden of proof under the AI Liability Directive comes into play, the challenged company must positively prove that it is not responsible for the damage caused.
• Disclosure obligations: Under the AI Liability Directive, injured parties should be entitled to demand disclosure of certain information from the operator, manufacturer or user of high-risk AI, as well as from distributors or other third parties subject to corresponding obligations under the AI Liability Directive.This information shall include training and validation data, details from the technical documentation and record keeping requirements, as well as information from the quality management system and on corrective actions taken. Disclosure of this information may only be made to the extent necessary and appropriate to pursue the claim. If disclosure of the information is wrongfully refused, it shall be presumed that the information requested would have justified the claim.
• Documentation requirement: Thorough and comprehensive documentation will become necessary for companies and AI users to effectively protect themselves from potential recourse claims. This is likely to pose particular challenges for small and medium-sized enterprises (SMEs), as they may not have the resources and expertise to meet all the requirements of the AI Liability Directive.