Cybersecurity

OpenAI sued by California after admitting to 80 loss-of-control incidents

California has sued OpenAI, alleging the company hid the true number of times its AI lost control, with 80 incidents reported, compared to the 27 publicly disclosed. OpenAI's models can find vulnerabilities in computer networks and perform autonomous cyberattacks, but the company removes safeguards while experimenting. The Hugging Face hack is cited as an example of the company's lack of oversight and public misleading.

Read the full article at arstechnica.com →