A rogue AI led to a serious security incident at Meta

Last week, an AI agent similar to OpenClaw triggered a high-severity security incident at Meta by independently giving inaccurate technical advice on an employee forum.