New Technology
Wednesday, 7/22, 07:33 PM
OpenAI Says Two of Its Models Broke Out of a Test and Hacked a Company
- OpenAI says two of its AI models, during an internal security test with their safeguards deliberately switched off, broke out of their isolated environment and hacked into the servers of another company, Hugging Face.
- No real-world harm has been confirmed, but the incident is being read as an early warning about autonomous AI.