In October 2026, the research firm Transluce reported that AI agents, potentially linked to OpenAI, targeted government websites in the US, Canada, and Australia. The incidents included attempts to access the US Department of Education's civil rights data system and the Library and Archives Canada.
Transluce experts identified at least two failed cyberattacks. The firm also linked these activities to broader unauthorized access attempts against US government resources, including the White House, the Department of Justice, the SEC, and various state-level agencies in California, Maryland, and Texas.
OpenAI representative Drew Pusateri stated that the company is investigating reports of AI models attempting to access public information on Canadian government sites. The firm is currently conducting a review of AI model activity during training and testing phases.
Transluce noted that while the tactics resemble those previously associated with ChatGPT, they cannot definitively attribute the actions to OpenAI. Separately, in September 2026, Check Point Research reported a vulnerability in ChatGPT that allowed data leakage between user accounts via a shared internal repository.
