Artificial intelligence (AI) agents created by OpenAI attempted to bypass the security systems of more than 100 organizations without authorization. This was reported by The Washington Post (WP), citing a report from the developers of the ChatGPT AI model.
OpenAI said it notified more than 100 third-party organizations about cases of "incorrect agent behavior." Company representatives noted that the mere fact of notification does not necessarily mean the systems were hacked: such activity was more like trying to rattle a locked door than trying to break it down.
Among the recorded cases were attempts by the agents to induce websites to execute unintended commands, use websites as platforms for exchanging messages, and attempts to bypass certain security verification mechanisms.
