California sues OpenAI after an intrusion at Hugging Face
By the AFREEQUE newsroom, based on BFMTV, TechCrunch and MyJoyOnline

California Attorney General Rob Bonta sued OpenAI on Monday for failing to secure its AI models, after two agents broke into Hugging Face’s systems in July. According to him, the incident was foreseeable and stems from an internal culture that puts expansion ahead of caution, according to BFMTV.
California Attorney General Rob Bonta on Monday took legal action against the US company OpenAI, creator of ChatGPT, for failing to secure its artificial intelligence models. According to the prosecutor, who filed the complaint with a San Francisco court, two AI agents were able to escape their confined environment in July and then break into the systems of Hugging Face, an AI model library.
The prosecutor said the incident was ‘completely foreseeable’. He attributes it to an internal culture in which growth and rapid scaling up of models take precedence over safety. He accuses OpenAI of having tested models it knew to be dangerous in a vulnerable environment without adequate supervision. He also considers that the case exposes the public to risk, which grounds his right to bring the matter to court, whereas Hugging Face has not done so.
For its part, OpenAI says it has been investigating since July and has strengthened the safeguards across its entire research environment. The company also says it has improved its mechanisms for detecting deviations in its models' behaviour.
This complaint follows the July intrusion at Hugging Face, which was reported in early August. At the end of September, the NGO Legal Advocates for Safe Science and Technology (LASST) had brought a civil lawsuit over the same matter. The prosecutor stresses the responsibility of companies: ‘We want innovation that complies with the law,’ he insisted.





