OpenAI agents accessed US government websites without authorization, following similar incidents in Australia

OpenAI has disclosed that its AI agents accessed US government websites without authorization after the systems operated in unintended ways.

According to the Associated Press, OpenAI said Friday (local time) that an internal review found its AI models had accessed two websites operated by the Securities and Exchange Commission and data held by the Census Bureau.

However, OpenAI said it found no evidence that the agents used SEC credentials, accessed non-public information, altered any data or systems, compromised security, or exploited vulnerabilities.

In a statement, OpenAI said it "continuously monitors for so-called alignment failure activity, where AI systems operate in unintended ways," adding that it "immediately notifies the relevant agencies when it identifies potential impacts on other systems."

OpenAI CEO Sam Altman also said Friday on social media that "a broad and ongoing investigation is underway into internet access that occurred during the training and evaluation of AI agents."

This is not the first time OpenAI's AI agents have attempted to access government sites. According to Australia's public broadcaster ABC, OpenAI agents recently attempted to access health statistics data from the Australian Department of Health as well as data from several other government agency websites, including the Australian Institute of Health and Welfare. In response, the Australian government said it would form a task force to investigate potential legal violations.