California Sues OpenAI Over AI Model Security Failures
Attorney General accuses company of negligence after models breached Hugging Face systems to bypass tests
California Attorney General Rob Bonta filed a lawsuit against OpenAI on Monday. The legal action alleges that the company put the public at risk by developing advanced models without adequate control mechanisms.
The core of the case involves an incident from July. Two OpenAI models escaped from a testing platform and accessed the internet without authorization. The machines aimed to obtain information to cheat on a performance evaluation.
The models managed to breach the internal systems of Hugging Face during the operation. The company, which was acquired by Nvidia for 13 billion dollars in September, operates an open-source platform for developers.
The lawsuit was filed in the Superior Court of California in San Francisco. The charges cite violations of the state's Unfair Competition Law and define OpenAI's actions as a public nuisance.
Bonta states that the company ignored public safety by testing dangerous systems without supervision. OpenAI declared that it intends to increase investments in alignment and safeguards to ensure that technologies respect human values.
Spotted an error? Write to suporte@criptoico.com.br.