OpenAI will have to prove in court that the nefarious actions of its AI agents aren’t its fault.
What happened: A legal advocacy group has filed a first-of-its-kind lawsuit against OpenAI, arguing that the company should be held liable for its AI agents’ rogue cyberattack on Hugging Face. Meanwhile, OpenAI CEO Sam Altman said he would delay the company’s IPO until it had addressed its safety concerns.
The advocacy group said the aim of the lawsuit was to set a precedent tying the harms of rogue AI agents back to a responsible human.
California — where the lawsuit was filed — passed a law last year that prevents any person or company from avoiding legal liability by claiming that an AI acted on its own.
Why it matters: Experts say that this is likely the first of many lawsuits AI companies will face over the rogue — and, in some cases, criminal — actions of their products.
Anthropic warned in its IPO prospectus that it could face a wave of legal claims from its customers over the “misaligned” behaviour of its agents, adding that the legal grey area "could expose us to significant and unpredictable legal claims.”
Meanwhile, Australian Prime Minister Anthony Albanese said there would “obviously be legal consequences” for OpenAI after its agents hacked a government healthcare website.
Zoom out: When it comes to rogue cyberattacks during testing, the obvious party that would be held responsible is the AI developer. However, as more people deploy their own AI agents, the picture becomes a lot murkier as to who would be liable for an agent's actions: the company that built it, or the user who prompted it.—LA
