European directive includes AI in product liability

A new European directive, effective December 9, 2026, holds companies responsible for damages caused by their AI tools. This means that companies, not the AI itself, will be liable for any harm caused by faulty software or artificial intelligence.

France adapts civil code to the directive

France is expected to update its civil code to match the new directive, but the exact timeline for these changes has not been announced. The directive applies to products placed on the market or put into service after the specified date.

"Casseur-payeur" principle debated in France

The "casseur-payeur" ("breaker-payer") principle has been discussed in French political debates following the violence of October 17, 2026. A bill was introduced to the National Assembly on October 13, 2026, and the government is defending its own version of the text, which the Senate will review starting October 27, 2026.

Court case shows AI liability in Canada

In February 2024, a court in British Columbia ruled that Air Canada was responsible for the actions of its chatbot, which gave incorrect information to a customer. The court rejected the company's argument that the chatbot was a separate entity and ordered Air Canada to pay about 812 Canadian dollars.

AI damage can be immediate or long-term

AI-related damage can be both immediate and gradual. Immediate damage includes errors such as incorrect information or misdirected emails. Gradual damage involves decisions based on AI without proper evidence, such as when Klarna admitted in 2025 that its automated customer service was cheaper but of lower quality, leading to the need to rehire human staff.

AI errors affect small businesses differently

In small and medium-sized enterprises, the financial impact of AI errors may not result in formal legal judgments. Instead, it may appear as lost customers, unspoken corrections by employees, or delayed discoveries by managers. The European directive aims to clarify that AI tool publishers can be held responsible like manufacturers.

Contractual agreements still matter

The new directive does not eliminate disputes between companies, as contractual agreements will still influence liability. Companies should review the liability clauses in their AI tool contracts, as these often limit the publisher's liability to a few months of subscription, leaving the using company to bear the risk.

Three key questions for AI use

Before entrusting a client to an AI system, companies should consider three key questions: who will fix the mistake and how, what the contract says about the publisher's liability, and whether the company can prove the AI's actions. These considerations help clarify the responsibilities and potential costs of using AI.