Commission publishes AI liability rules

When your toaster or kettle fails, the product liability laws you rely on are pretty well established. When you or your property are harmed by an errant delivery drone, it's a different matter altogether. 

It's in this context that the EU Commission has just published its draft AI Liability Directive, and it's likely to make for nervous reading across the boardrooms of companies developing or deploying products or services reliant on AI.

Key principles include:

• a presumption of causality where fault is clearly established, meaning victims would not need to describe in detail how the harm was caused - something the average consumer would likely struggle to do when dealing with a complex AI system; an
• an enhanced right of access to evidence from companies and suppliers, which could mean companies being required to disclose information about the design or functionality of its AI systems to allow liability to be fully assessed.

These principles do build on themes of accountability, transparency and 'explainability' common across many of the draft AI regulations that are emerging across the globe, including the UK, but they add a whole new layer of expectation around the remedies that consumers will have at their disposal where AI fails.

The Directive will sit alongside the EU's AI Act which is due to come into effect next year, but in the meantime companies looking to develop or supply AI into EU countries, particularly in 'high risk' applications or environments, would do well to start planning for an enhanced liability regime.Our team of Shoosmiths AI specialists will be monitoring developments on both pieces of legislation carefully, alongside the UK Government's latest proposals, as they proceed towards implementation.