Gray Swan

Pittsburgh, Pennsylvania, USA · ≈97 employees · grayswan.ai

Gray Swan is an AI security company founded in 2023 by researchers from Carnegie Mellon University. The company specializes in automated adversarial testing, continuous red teaming, and runtime AI security to protect AI models and agents in production. Gray Swan serves frontier AI labs, such as Anthropic, OpenAI, and Meta, as well as enterprise organizations that deploy AI agents and require real-time protection against vulnerabilities. The company's platform includes three core components: Arena, a large adversarial AI red teaming network; Shade, a continuous adversarial testing tool that identifies vulnerabilities; and Cygnal, a real-time AI firewall that blocks threats in production. Gray Swan also offers Gray Swan Cygnet, a secured version of Meta's Llama 3 model designed to be less susceptible to adversarial attacks. With a mission to empower safe and secure AI usage, Gray Swan aims to assess and mitigate unintended behaviors in AI systems.