Faramir
Faramir makes it safe for the world to give AI real responsibility.
Year
2026-Present
Project overview
Faramir makes it safe for companies to let AI run real parts of their business. Our safety harness tests internal AI systems before deployment, then monitors and stops them in production when they access the wrong data, misuse tools, or act beyond their authority.
Project process
Conducted user discovery calls with 50+ companies in biotech, fintech, healthcare, and tech where we validated the scale of this issue. Learned OpenAI's model went rogue and hacked a competitor, yet there are not safeguards in place in every company we talked to preventing such incidents from reoccuring. Brought on experts as advisors including U.S. National Science Foundation, Cerebras, Lockheed Martin, Crowdstrike, and two cybersecurity experts ranked 1 and 2 nationally on CTFtime.org respectively.
Final result
We launched customized models we are now preparing to deploy custom solutions with our customers. Since starting, we have received immense industry industry, alongside paid pilots. We are profitable and currently finalizing a $50K contract with a customer.