Gray Swan is an AI security company spun out of Carnegie Mellon University. It helps frontier AI labs and enterprises find and fix dangerous vulnerabilities in their AI models before and after deployment. The company operates three products: Arena, a crowdsourced red-teaming network of 15,000+ security researchers who actively try to break frontier models like GPT-5, Claude, and Gemini; Shade, an automated adversarial testing agent that continuously probes enterprise AI systems using attack patterns gathered from Arena; and Cygnal, a runtime protection layer that monitors and blocks malicious inputs and outputs in production. Customers include every major frontier lab — OpenAI, Anthropic, Google DeepMind, Meta, xAI, and ByteDance — as well as enterprises building on AI infrastructure, such as Snowflake. Gray Swan has been cited in 11 frontier model system cards, meaning labs publicly credit its work in their official safety disclosures. The company is now pushing aggressively into enterprise sales alongside its established frontier-lab business.