European lawmakers split over AI safeguards and liability

The European Parliament backed deeper international cooperation on AI governance during a Strasbourg debate, but lawmakers disagreed over regulation and corporate responsibility. Proposals included a binding agreement involving the United States and China, an arms-control-style treaty and an international inspection regime requiring testing before systems reach the market.

Greens and Socialists rejected industry self-regulation, warning that financial incentives can undermine oversight. Swedish center-right lawmaker Arba Kokalari countered that additional rules risk weakening European developers already trailing American and Chinese competitors. German lawmaker Sergey Lagodinsky argued that reports of autonomous agents escaping testing environments and increasingly sophisticated hacking capabilities justified stronger safeguards.

The debate also renewed pressure on the European Commission to revive the withdrawn AI Liability Directive, intended to help people seek compensation for AI-related harm. Commissioner Jozef Síkela defended the withdrawal, citing resistance from national governments and little prospect of agreement. He said the AI Act and revised Product Liability Directive already address some liability issues. Critics argued that those measures leave gaps: regulatory fines do not necessarily compensate victims, while proving fault and causation remains difficult when AI systems make autonomous decisions.