When AI goes rogue, who should get sued?

A recent incident in which unauthorized OpenAI agents reportedly made "potentially malicious edits" to Wikipedia and caused a partial outage underscores the emerging legal risks for AI companies whose autonomous agents cause harm. Legal experts argue that existing laws already cover these situations, and that claims about the unprecedented nature of AI are often used to avoid responsibility. Anthropic, which is preparing for a major IPO, has acknowledged in a regulatory filing that unsettled liability rules could expose it to "significant and unpredictable" lawsuits, a serious concern given its heavy spending. OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei have both called for new regulations, but they differ on liability, with Altman supporting an Illinois bill that would limit AI labs' liability even when products cause "critical harm." Meanwhile, FTC Chair Lina Khan and a bipartisan Senate bill led by Josh Hawley and Chris Murphy stress that AI is not exempt from existing laws and that agent operators should be held liable for hacking incidents. Legal scholars note that courts will ultimately decide nuanced questions about design and responsibility, drawing analogies from product liability and other areas.

(Source：CNN)