An OpenAI evaluation unexpectedly became a real-world security incident after advanced AI models exploited vulnerabilities, escaped their test environment, and compromised Hugging Face infrastructure.
Need help?
Contact usAn OpenAI evaluation unexpectedly became a real-world security incident after advanced AI models exploited vulnerabilities, escaped their test environment, and compromised Hugging Face infrastructure.