BRUSSELS — Any company operating within the European Union that deploys AI to make decisions about hiring, lending, criminal justice, healthcare, or any domain deemed "consequential to individual rights" must now carry mandatory liability insurance, maintain auditable decision logs, and have a named human accountable for every algorithmic outcome. The AI Liability Directive, passed 571-46 by the European Parliament in late 2025, is now enforceable across all 27 member states.

The directive shifts the burden of proof: individuals claiming harm no longer need to prove the AI system was defective. Instead, the deploying company must prove its system was not at fault — a "rebuttable presumption of causality." "That is a 180-degree reversal from the previous state of affairs," explained Professor Anja Hoffmann of the University of Amsterdam.

Companies deploying high-risk AI must maintain coverage of at least €10 million per incident. Major European insurers including Allianz and AXA have launched dedicated AI liability products, though premiums remain high at 1.5-4% of coverage annually, reflecting legal uncertainty. American tech giants Google, Microsoft, and OpenAI have each established EU AI compliance teams numbering hundreds of employees. Amazon has suspended its AI recruitment tool across EU operations pending an 18-month compliance review.