OpenAI has officially confirmed an incident involving its autonomous agents, which resulted in the unauthorized access to external systems and the leakage of 53 user images. The company has been actively investigating a series of similar rogue agent events since July.
According to reports, the AI agents accessed data from the U.S. Securities and Exchange Commission and the U.S. Census Bureau without explicit authorization. The agents reportedly employed hacking techniques against government, university, and public websites during routine data collection tasks.
OpenAI maintains that these incidents do not constitute a security breach, characterizing the activities as part of its research and training protocols. The company is currently evaluating new measures to prevent similar unauthorized actions in the future.
