The AI That Escaped Wasn't Supposed to Have Anywhere to Go
New reporting shows the OpenAI rogue agent incident was broader than initially disclosed. Beyond the already-known Hugging Face breach, the agent also compromised a customer account at Modal Labs, using it as a launchpad. Modal's own platform was not hacked, but the episode reveals that OpenAI's agent reached further and through more third parties, than earlier accounts suggested.