律动BlockBeats
律动BlockBeats|Nov 17, 2026 03:34
[Over 80 AI Out-of-Control Incidents, Only 27 Publicly Disclosed! California Sues OpenAI, Demanding Court-Mandated Rectification] Beating AI News Flash: On November 16th local time, California Attorney General Rob Bonta, where OpenAI is headquartered, formally sued OpenAI, accusing it of knowingly conducting training of experimental AI models in an environment lacking adequate security safeguards. This is the first time a US state government has filed a lawsuit against an AI agent for going out of control and attacking external systems. The indictment reveals that OpenAI admitted during the California investigation that it lost control of its AI agents more than 80 times during testing, but only publicly disclosed 27 of these incidents. The California prosecutor also alleges that OpenAI refused to provide detailed records of the remaining incidents, preventing regulators from assessing the true risks. In July of this year, OpenAI tested its AI in an isolated environment where these agents were not allowed to connect to the external network. However, they exploited vulnerabilities in internal software services, exchanged information with each other, and broke through the isolation. Approximately 700 agents subsequently attacked the AI open-source platform Hugging Face, launching about 17,600 attacks in two and a half days, forcing the Hugging Face security team to spend two full weeks dealing with the situation. That same month, another group of experimental agents broke into OpenAI's internal network, even gaining administrator privileges on critical systems such as cloud model deployment, identity authentication, and confidential credential management. California prosecutors believe these incidents could have been avoided. OpenAI discovered as early as May that a model had breached its isolation environment and connected to the internet without authorization, yet it did not stop training. After the July incident, the company strengthened security measures, but in September, models again breached the isolation. Although the monitoring system detected the anomaly within minutes, training did not automatically stop; manual reviewers took approximately two hours to halt it. Last year, the California Department of Justice participated in overseeing OpenAI's corporate restructuring and reached a security governance agreement requiring the company to prioritize security, with the security committee even having the power to halt the release of new models. Now, the California Attorney General, responsible for oversight, has directly sued OpenAI. California has accused OpenAI of violating unfair competition laws and constituting "public nuisance" that endangers public safety. Prosecutors are asking the court to order the elimination of the ongoing safety risks, prohibit the illegal activities, recover illicit gains, and impose civil penalties. If granted an injunction, OpenAI may be forced to adjust its model training and testing processes as required by the court. OpenAI responded that it has launched an investigation, strengthened security measures, and notified affected organizations following the Hugging Face incident. The company believes that uncontrolled intelligent agents are a challenge faced by the entire AI industry, and that this lawsuit targeting OpenAI alone is unfair. [Original Link]
+4
Mentioned
Share To

Timeline

HotFlash

APP

X

Telegram

Facebook

Reddit

CopyLink

Hot Reads