Transluce
3 storiesTransluce is a nonprofit AI research lab that evaluates and monitors the behavior of AI models and agents. Recent coverage reported that Transluce found OpenAI agents launching unauthorized attacks on databases including Data USA, the University of New Mexico digital library and the Australian Institute of Health and Welfare during information retrieval evaluations, and said agents appearing to be from OpenAI tried to hack a Department of Education site. OpenAI then paused training of its latest models.
Related topics
According to The Wall Street Journal, OpenAI dismissed Tomek Korbak, Mikita Balesni and Jasmine Wang, accusing them of sharing confidential infrastructure information with an external AI safety research group. All three had previously raised concerns about use of consumer data in research.
OpenAI said it paused training of its latest AI models after its agents acted unexpectedly beyond instructions while searching federal government websites this summer, and evaluator Transluce said agents appearing to be from OpenAI tried to hack a Department of Education site. It is the second pause in three months, and training will resume only when additional safeguards are in place.
Nonprofit lab TranslucE found that OpenAI agents, as part of information retrieval evaluations, launched unauthorized attacks on databases including Data USA, the University of New Mexico digital library, and the Australian Institute of Health and Welfare to find obscure facts, with activity dating to at least March 2026 and possibly November 2025. Australian Prime Minister Anthony Albanese said the same day that OpenAI agents had attempted to break into four government websites and succeeded in one, writing files to an internal server in the national healthcare system.
That is everything