On a quiet stretch of Ellsworth Avenue in Shadyside, the team at Gray Swan is doing something that most of the world's largest technology companies are paying close attention to: figuring out how to stop AI from going wrong. This past spring, the Pittsburgh startup announced the close of a $40 million Series A — the largest funding round yet for a company founded just three years ago on research born inside Carnegie Mellon University's legendary School of Computer Science.
The round was co-led by Wing Venture Capital, the California-based firm that backed Snowflake and HashiCorp, and Madrona, the Seattle investor behind Amazon Web Services' earliest institutional support. Additional participation came from Obvious Ventures, Snowflake Ventures, Hudson River Trading, Samsung Next, and Pittsburgh's own Magarac Venture Partners — a roster that reads like a who's-who of the institutions shaping the future of computing.
"AI applications are growing at an unprecedented rate, and at Gray Swan, we want to ensure that these deployments can continue without sacrificing reliability and security."
Zico Kolter, Chief Scientist and Co-founder, Gray Swan
Gray Swan was co-founded by Matt Fredrikson, who serves as CEO, and Zico Kolter, the company's chief scientist. Both built their reputations at CMU, where they spent years studying the ways AI models can be manipulated, tricked, or made to behave in dangerous ways. That research — unglamorous but essential — turned out to be exactly what the technology industry needed as large language models began powering enterprise applications, healthcare systems, financial platforms, and autonomous agents at scale.
Three Layers of Defense
The company's platform operates across three integrated products. Shade tests AI models for vulnerabilities before they ever go live, running automated adversarial attacks to find weaknesses in the kind of controlled environment that prevents those weaknesses from being discovered in the wild. Cygnal watches deployed AI systems in real time, blocking dangerous outputs, prompt injection attempts, and data exfiltration before they can cause harm. And Arena — perhaps the most audacious piece of the stack — is a global network of more than 15,000 ethical hackers who probe AI models on a continuous basis, feeding fresh threat intelligence back into both Shade and Cygnal.
The result is a security ecosystem designed not for a static world, but for one where AI capabilities are expanding month by month. Gray Swan counts over 20 customers across frontier AI labs and global enterprises, and its work has been cited directly in safety reports published by Anthropic, OpenAI, and Meta — the three organizations leading the development of the world's most capable AI systems.
Pittsburgh's AI Security Capital
The new funding will go toward accelerating Gray Swan's go-to-market operations, deepening its existing partnerships with frontier labs, and scaling the team. The company is actively hiring across engineering, sales, and marketing, with open roles carrying salaries up to $260,000 — a sign that Shadyside is becoming a destination for elite AI security talent that might otherwise default to San Francisco or New York.
"Our mission is to empower the world to use AI safely and securely, and this funding lets us pursue the mission on pace with the frontier," Fredrikson said in the announcement. The pace he is referring to is genuinely staggering: the AI capabilities Gray Swan is asked to protect against are themselves evolving, which means the company is building a moving target into its core product design.
Gray Swan was ranked first on Technical.ly's 2026 RealLIST Startups in Pittsburgh — a recognition that reflects both the company's trajectory and the broader moment this city's tech ecosystem is experiencing. Pittsburgh attracted $1.7 billion in startup investment in Q1 of this year alone, and companies like Gray Swan represent the cutting edge of that wave: deeply research-rooted, globally relevant, and built right here on Ellsworth Avenue.
For a city that has spent decades reinventing itself after the steel era, the emergence of a company whose job is literally to make artificial intelligence trustworthy carries a particular kind of resonance. Pittsburgh trained the researchers who built the tools that power the modern AI economy. Now it is training the researchers who will keep that economy safe.