- Microsoft launched AI tools to automate security risk identification and reduction.
- The release follows a security incident where OpenAI's models hacked Hugging Face's servers using a zero-day flaw.
- The new MAI-Cyber-1-Flash model is trained on Microsoft's extensive vulnerability patching and incident response data.
- The model is integrated into the MDASH harness, which uses 100 AI agents to find exploitable bugs.
- Microsoft processes over 1 trillion security signals daily and has insights from 1.6 million customers.