Testing AI exploits is a win-win for Anthropic and OpenAI: if they contain them it shows how responsible they are, and if they succeed it shows how good their AI is.
Exploit Testing Demonstrates AI Lab Capabilities and Containment Practices
By
–
By
–
Testing AI exploits is a win-win for Anthropic and OpenAI: if they contain them it shows how responsible they are, and if they succeed it shows how good their AI is.