CMU / Zico Kolter
Providing enterprise-grade security solutions for LLMs and AI-powered applications through cutting-edge risk assessment and mitigation tools.
Founding Story
Founded by the researchers who discovered the GCG jailbreak method, which bypassed safety filters on all major LLMs. The team launched Gray Swan AI to transition their academic discoveries into a comprehensive security suite for enterprises and AI labs.
Discussions
No discussions yet
Be the first to start a discussion about CMU / Zico Kolter
Leadership
Founders
Zico Kolter
Professor and Director of the Machine Learning Department at Carnegie Mellon University; OpenAI Board Member; Qualcomm Board Member; previously Chief Expert at Bosch.
Matt Fredrikson
Associate Professor of Computer Science at Carnegie Mellon University; expert in security and privacy in machine learning.
Andy Zou
Researcher at Carnegie Mellon University; lead author of the 'Greedy Coordinate Gradient' (GCG) jailbreak paper.
Executive Team
Matt Fredrikson
Co-founder & Chief Executive Officer
CMU CS Professor and expert in ML security.
Zico Kolter
Co-founder & Chief Scientist
CMU ML Dept Director and OpenAI/Qualcomm Board Member.
Business Model
Revenue Model
B2B SaaS and API-based usage model for security services and enterprise subscriptions.
Pricing Tiers
Full access to Security Suite, Cygnal, and Shade with API support.
Free/community access for researchers and bounty programs.
Target Markets
- Enterprise companies deploying AI
- Frontier AI labs
- Model developers
- Security researchers
- Enterprise AI security for autonomous agents
- Protecting tool/database/API integrations (MCP)
- Model safety evaluation for AI labs
- Regulatory and compliance testing
- AI red-teaming career training (Proving Ground)
- OpenAI
- Anthropic
- US AI Safety Institute
- UK AI Safety Institute
History & Milestones
Appointed Rob Jenks as Chief Strategy Officer.
US AI Safety Institute joins Gray Swan's Agent Red-Teaming Challenge as co-judge.
Introduced Gray Swan AI Proving Ground for training AI red-teamers.
Published study with Stanford showing AI agents outperform humans in penetration testing.
Emerged from stealth with official public launch.
1 AI Tool by CMU / Zico Kolter
A free online course by CMU's Zico Kolter covering the machine learning methods and large language models behind modern AI systems like ChatGPT, Gemini, and Claude.
