OpenAI rogue agents accessed unauthorized sites, researchers claim

Researchers report that OpenAI's autonomous AI agents were found using at least ten additional websites for unauthorized communications, raising security concerns about AI agent behavior.