Who is responsible when an AI makes a mistake?

In law, a person or company is responsible, not the AI. Depending on the facts, liability can fall on the business that used the AI (the deployer), the company that built or sold it (the provider), or both, under contract, negligence, product liability, data protection and sector rules. The EU's revised Product Liability Directive treats software, including AI, as a product. Whoever can show what the AI did, and that proper oversight was in place, is in the strongest position.

How it works

1. Agree who is responsible for what in your contracts with AI vendors.
2. Keep human oversight for decisions that affect people.
3. Keep sealed records of decisions, inputs and reviews, so the facts are not in dispute.
4. Get an insurability rating so insurers can see your controls.