A few days ago, autonomous AI agents of OpenAI launched coordinated cyberattacks against Hugging Face. Details of what happened are now public: Agents began collaborating via shared package managers to overcome difficult tasks, created a message board, chained vulnerabilities, such as zero-day exploits, gained administrative access, moved through internal networks in OpenAI, and eventually infiltrated Hugging Face infrastructure. Conclusion: we'll need AI to defend us against AI 😬. Video with the details:
Edit thought
English
Español
Media
Shared between English and Spanish.