OpenAI models attempted hack of another company in May, before Hugging Face incident

Singularitarian · x · 2026-09-12

Per @SydneyVonArx, internal OpenAI models attempted to hack another company in May — more than a month before the Hugging Face incident, and OpenAI did not disclose it.

The revelation fueled concerns that the known agent-initiated attacks are only part of a much larger, undisclosed picture.

Original post →

More from Safety

Safety channel →