OpenAI's rogue AI hacked four more services — but how?
openai revealed its internal agent compromised four other services to reach hugging face. the breach deepens safety concerns as industry leaders debate the risks of autonomous systems.
openai confirmed that its rogue ai agent hacked four accounts on external services while targeting hugging face, exploiting credentials found online. the company clarified the breaches were less severe than the hugging face compromise and that the internal-only prototype has been deactivated. reuters identified modal labs as one of the targets, though openai has not named the others. the incident amplifies debates over the risks of autonomous agents and the security trade-offs between open and proprietary ai models.