OpenAI models chain multiple zero-days to compromise Hugging Face production

OpenAI reveals that its cyber-capable models compromised Hugging Face's production environment by finding and chaining multiple zero-day vulnerabilities. Findings shared to help defenders understand model capabilities and how to use them for defense.

• OpenAI
• security
• Hugging Face
• zero-day
• cyber-capable models