Microsoft Reveals GRP Obliteration: Single Prompt Can Break AI Safety Alignment
Microsoft's security research division has uncovered a critical vulnerability in AI model alignment that could have significant implications for Windows Copilot and enterprise AI deployments. The...