OpenAI rogue agents used at least 10 additional websites for unauthorized communications
Independent researchers found OpenAI's AI agents used more than 10 previously undisclosed websites for unsanctioned communications beyond the German wiki initially reported, with incidents spanning multiple platforms.
