When the largest AI laboratories in the world need someone to stress-test their models — to probe them for vulnerabilities, trick them into dangerous outputs, and find every edge case before it becomes a real-world problem — a growing number of them have turned to a company on Ellsworth Avenue in Shadyside. Gray Swan, the Carnegie Mellon University spinout that has quietly become the most credentialed AI safety testing firm in the country, announced the close of a $40 million Series A round in late May, its largest financing to date and a clear signal that the market for AI security has arrived in force.

The round was co-led by Wing Venture Capital, based in California, and Madrona, the Seattle-based firm that has backed some of the Pacific Northwest's most durable technology companies. Additional participation came from Obvious Ventures, Snowflake Ventures, Hudson River Trading, Samsung Next, and Pittsburgh's own Magarac Venture Partners, which has backed Gray Swan since the beginning. The company declined to disclose its valuation.

"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 founded in 2023 by Matt Fredrikson, who serves as CEO, and Zico Kolter, chief scientist, both of whom built their reputations as researchers at CMU studying the ways AI systems fail, break, and can be manipulated. That academic grounding has translated directly into product: the company's platform has three components operating in concert. Shade tests AI models for weaknesses before they are ever deployed. Cygnal monitors live deployments in real time, blocking dangerous behavior as it emerges. And Arena, perhaps the most distinctive element, is a global network of more than 15,000 ethical hackers who actively probe AI models for vulnerabilities and feed their findings back into both tools, keeping Gray Swan's threat intelligence current as models evolve.

The result is a company that now works with more than 20 customers across frontier labs and major global enterprises — and whose work is cited in the public documentation of the companies widely considered to be leading the AI race. Anthropic, OpenAI, and Meta have all referenced Gray Swan's research in their published model cards and safety reports, a form of third-party validation that most startups spend years trying to earn.

Gray Swan at a Glance
$40M Series A round closed May 2026, led by Wing Venture Capital and Madrona
15,000+ Ethical hackers in Gray Swan's Arena network, continuously probing AI models for vulnerabilities
20+ Enterprise and frontier lab customers, including Snowflake, which also invested in the round
#1 Ranked top Pittsburgh startup by Technical.ly's 2026 RealLIST Startups report

Building the Security Layer for an AI-First World

The timing of Gray Swan's raise is no coincidence. As artificial intelligence moves from research labs into hospitals, banks, legal firms, and government agencies, the question of what happens when a model behaves unexpectedly has shifted from an academic concern to a commercial liability. Companies deploying AI at scale need to know their systems won't leak sensitive data, generate harmful content, or be manipulated by adversarial inputs. Gray Swan is building the infrastructure to answer those questions — before the incident, not after.

Fredrikson said the company plans to use the new capital to accelerate its go-to-market operations, deepen its partnerships with frontier labs, and scale its team across engineering, sales, and marketing. The company is currently hiring for all three areas, with technical roles offering salaries reaching $260,000 — a figure that reflects both the difficulty of finding people who deeply understand AI security and the company's ambition to compete for talent at the highest level of the market.

A Pittsburgh Success Story With Deep Roots

Gray Swan's trajectory is the story Pittsburgh's technology community has been telling about itself for years, now made concrete. The company emerged from foundational research conducted at Carnegie Mellon, one of the world's foremost institutions for artificial intelligence. It built a product grounded in that research. It raised early money from Magarac Venture Partners, a Pittsburgh firm that has made a deliberate bet on the city's AI ecosystem. And it stayed, planting its headquarters in Shadyside rather than relocating to the Bay Area or New York as so many funded startups eventually do.

That staying power matters. Gray Swan is now one of the most closely watched AI companies in the country, and it is watching the world from Ellsworth Avenue. Pittsburgh's technology sector — anchored by CMU and the University of Pittsburgh, shaped by decades of robotics and autonomous systems research, and increasingly recognized for producing companies that can compete at the highest level — has in Gray Swan a company that embodies what the ecosystem is capable of producing. The Series A does not end that story. It is, if anything, the moment it properly begins.