“Stress test” platform serving the major AI frontier models, including Anthropic, OpenAI, and Meta, secures $40M in Series A funding.
The company closed a $40 million Series A co-led by Wing Venture Capital and Madrona. Obvious Ventures, Snowflake Ventures, Hudson River Trading, Samsung Next, and existing investor Magarac Venture Partners also participated.
Gray Swan builds security tools for AI systems. Gray Swan helps enterprises and AI labs find vulnerabilities in AI models and agents before deployment, then block dangerous behavior in real time. Its platform has three components: Cygnal, which monitors AI outputs and enforces custom security policies set by the organization; Shade, which automatically runs adversarial tests against AI deployments before they go live; and Arena, a global competition where more than 15,000 security researchers probe AI models for weaknesses. Arena has generated more than 1 million attack trajectories that feed the threat intelligence for Cygnal and Shade.
Gray Swan’s adversarial evaluations are cited in 11 recent frontier model system cards, including those from Anthropic, OpenAI, and Meta. System cards are the safety documents labs publish before releasing a model to the public; Gray Swan’s testing is embedded in that process. The company says no other AI security vendor holds that position across all three major labs simultaneously. It works with more than 20 customers across frontier labs and global enterprises.
Snowflake partnership
Gray Swan also announced a partnership with Snowflake. Its real-time protection now runs inside Snowflake’s AI platform. Enterprises that build and run AI applications on Snowflake can apply Gray Swan’s security layer directly in that environment, without a separate integration step. Snowflake is one of the primary platforms where enterprises deploy AI on their own data, making the integration relevant to companies already in that stack.
The funding will go toward sales expansion, deeper lab partnerships, and team growth.
