‘UNPRECEDENTED’

Rogue AI Escapes Human Control Just Like We Feared

Elon Musk called the incident “Troubling …”

A rogue AI agent just achieved something researchers have long feared: It broke out of its testing environment, reached the internet, and hacked another AI company. OpenAI said Tuesday that one of its autonomous AI agents escaped a tightly controlled environment and compromised the infrastructure of AI startup Hugging Face last week. The company called it “an unprecedented cyber incident, involving state-of-the-art cyber capabilities.” It warned such incidents are “something we expect to become more commonplace with the proliferation of increasingly cyber-capable models.” OpenAI said the autonomous agent exploited vulnerabilities to escape its sandboxed testing environment before infiltrating Hugging Face’s systems. The AI was powered by a combination of OpenAI models, including GPT-5.6 Sol and a more capable pre-release model, with cyber refusals reduced for evaluation purposes. Hugging Face had disclosed last week that it was the target of a hack “driven, end to end, by an autonomous AI agent system.” OpenAI said it was investigating the incident alongside Hugging Face and would release more details when the probe is complete. Elon Musk, who co-founded OpenAI with Sam Altman and later unsuccessfully sued them, reacted to the incident on X, writing: “Troubling …” Musk now runs rival xAI.