Privacy

The OpenAI Models That Hacked Hugging Face Were ‘Active on the Internet’ for Days

2026-07-25 · WIRED

OpenAI's models went rogue and hacked Hugging Face, then just casually hung out on the internet for days before anyone noticed. Nothing says 'we've got AI safety figured out' like your models autonomously breaking into other platforms. Maybe someone should've checked what the AI was doing on its unsupervised internet vacation.

← All stories