Gray Swan funding, valuation and investors

Overview

Gray Swan is a Pittsburgh-based AI security company spun out of Carnegie Mellon University's AI safety research program. CEO Matt Fredrikson and chief scientist Zico Kolter built a platform covering the full AI lifecycle: Shade runs automated adversarial tests in CI/CD, Cygnal blocks adversarial inputs and unsafe outputs at runtime with custom policies, and Arena hosts 15,000+ ethical hackers generating 1M+ real-world attack trajectories that continuously retrain defenses. The company serves 20+ global enterprises and is embedded in pre-release safety evaluations at leading frontier labs. Gray Swan closed a $40M Series A in May 2026 co-led by Wing Venture Capital and Madrona, with Snowflake Ventures, Obvious Ventures, Hudson River Trading, and Samsung Next participating.

Why Gray Swan is interesting

Gray Swan's adversarial testing is cited in frontier model system cards from OpenAI, Anthropic, and Meta — then productized as Cygnal runtime protection and Shade automated red-teaming for enterprises.

Product & use cases

Three integrated products: Cygnal (inline runtime monitoring blocking adversarial prompts, unsafe outputs, and tool-call abuse), Shade (LLM-powered adversarial red-teaming in pre-deployment and CI/CD), and Arena (continuous global competition feeding live threat intelligence into both).

• Frontier AI labs running pre-release safety evaluations before public model launches
• Enterprises deploying AI agents with custom policy guardrails in production
• Security teams automating adversarial testing in CI/CD pipelines
• Snowflake customers extending AI security to enterprise data platforms via partnership

Key facts

• Series A (May 2026): $40M co-led by Wing and Madrona — company press release
• Cited in 11 recent frontier model system cards including Anthropic, OpenAI, and Meta (Fintech.Global)
• Arena community: 15,000+ researchers generating 1M+ attack trajectories
• Carnegie Mellon spinout led by CEO Matt Fredrikson and chief scientist Zico Kolter
• Snowflake partnership extends security to enterprise data platforms

Funding history (newest first)

Competitive landscape

Edge: Threat intelligence from 15K+ Arena researchers breaking current-generation models — not stale public attack datasets — plus credibility embedded in multiple frontier lab system cards simultaneously.

AI security is fragmenting between data-security vendors adding LLM guardrails and dedicated AI-red-team startups. Gray Swan's wedge is dual: trusted by frontier labs for evaluation (11 recent system card citations per company) and productized for enterprise runtime — a combination competitors rarely hold simultaneously.

• Lakera direct

  LLM firewall and red-teaming; lacks Gray Swan's frontier-lab evaluation pedigree and Arena community scale.

• Robust Intelligence (Cisco) direct

  AI security validation acquired by Cisco; enterprise-focused vs. Gray Swan's lab-grade research roots.

• HiddenLayer direct

  ML and LLM security platform; competes on runtime protection but without Arena-sourced live attack feeds.

• In-house lab safety teams alternative

  Frontier labs build internal red teams; Gray Swan sells external evaluation standardization and enterprise-grade tooling.

Notable stories

• Gray Swan originated from Carnegie Mellon's AI safety research team — the same academic lineage as chief scientist Zico Kolter's adversarial ML work cited across frontier lab evaluations.
• The company claims no other AI security provider is simultaneously embedded in pre-release safety evaluations at multiple leading frontier labs — a distribution moat competitors cannot easily replicate.

Industries

Related funding articles

Venture Capital Tracker pieces that cover Gray Swan's financing or category context.

• Gray Swan's $40M Series A: The AI Security Firm Trusted by Every Frontier Lab