"Microsoft's PyRIT: A Free GenAI Red Teaming Tool for Cybersecurity"
Microsoft has released PyRIT, an open access automation framework designed to proactively identify risks in generative artificial intelligence systems. The tool assesses the robustness of large language model endpoints against various harm categories and can identify security and privacy harms. It comes with five interfaces and offers options for scoring outputs from the target AI system. While not a replacement for manual red teaming, PyRIT complements a red team's domain expertise by highlighting risk "hot spots" and generating prompts for evaluation. Microsoft emphasizes the need for both manual probing and automation in red teaming generative AI systems, as Protect AI discloses critical vulnerabilities in popular AI supply chain platforms.