What happened
Gray Swan, an AI security company spun out of Carnegie Mellon University and co-led by CEO Matt Fredrikson and Chief Scientist Zico Kolter, announced a $40 million Series A co-led by Wing Venture Capital and Madrona on May 28, 2026. The company has been cited in 11 recent frontier model system cards (including Anthropic, OpenAI, and Meta), making it the only AI security firm with embedded access across all three leading frontier labs simultaneously. Gray Swan's platform (Shade for adversarial testing, Cygnal for real-time protection, Arena for community red-teaming) now integrates natively with Snowflake's AI ecosystem for runtime AI protection.
Why it matters
Gray Swan's unique position — the sole firm with evaluation trust from all three major frontier labs — means its threat intelligence and red-teaming methodology is shaping how the most advanced AI models are tested before deployment. The Snowflake integration brings that same adversarial rigour directly into the enterprise data platform where AI applications are built at scale, creating a novel capability for enterprises to embed pre-vetted AI security controls without building a separate evaluation programme.
Applicability
Organisations deploying AI agents or models that handle sensitive enterprise data in Snowflake should evaluate Gray Swan's Cygnal integration. CISO teams building AI red-teaming programmes should consider whether Gray Swan's frontier-lab-grade adversarial testing (Shade) delivers ROI versus building in-house evaluation. AI security consulting firms should track Gray Swan as a key third-party evaluation vendor for client engagements.