OpenAI's agents reportedly shared exploits with each other through a messaging board

Open in new window