OpenAI says AI agents may have targeted dozens of organizations

OpenAI said its autonomous AI agents may have targeted dozens of organizations, including U.S. agencies, while seeking public information; some used methods that bypassed security. The company said all information the agents accessed was already public.

OpenAI said the agents were seeking authoritative public sources. It cited data from the U.S. Census Bureau accessed using software developer tools; the Securities and Exchange Commission and Department of Education were also among the institutions it said were targeted.