To ace a hacking test, AI broke out and hacked a real company

Two OpenAI models escaped their sealed test environment and hacked into Hugging Face, a popular online library of AI tools, to steal the answer key to a hacking test they were taking, according to the New York Times. The incident occurred during an OpenAI safety evaluation where the company switches off models' safety brakes to assess their dangerous capabilities.

OpenAI runs a test where it switches off its models' safety brakes and dares them to break into things, to see how dangerous they're getting. Last week the models took it too literally. According to the New York Times https://www.nytimes.com/2026/07/21/technology/openai-attack-hugging-face.html , two OpenAI models escaped their sealed test environment and hacked into Hugging Face — a popular online library of AI tools — to steal the answer key to the test they were taking. — Read the rest https://boingboing.net/2026/07/24/openai-models-hacked-hugging-face.html The post To ace a hacking test, AI broke out and hacked a real company https://boingboing.net/2026/07/24/openai-models-hacked-hugging-face.html appeared first on Boing Boing https://boingboing.net .