AI
A 40 Million Dollar Bet That AI Agents Are Ready for Their Own SOC 2
The Artificial Intelligence Underwriting Company, a startup founded by Anthropic's first product hire Rune Kvist and former METR chief operating officer Rajiv Dattani, announced Tuesday a 40 million dollar Series A led by Ribbit Capital with First Harmonic participating. The round brings AIUC's total funding to 55 million dollars, following a 15 million dollar seed led by Nat Friedman's NFDG. The two founders, who are brothers in law, are betting that enterprises will pay for proof that AI agents behave.
AIUC's product is an independent audit and certification layer for AI agents, modeled on the cybersecurity standard SOC 2. The company built a standard called AIUC-1 together with a consortium of more than 250 security leaders from Fortune 1000 companies, the actual buyers of enterprise agents. Each agent runs through about 5,000 security and attack combinations covering jailbreaks, hallucinations and data leaks, producing a roughly 100 page report on where the agent performs reliably. AI agents run the tests and AI analyzes the data, while humans verify the final audit. Agents earn recertification every quarter.
The customer list is the tell. Cursor, ElevenLabs, Harvey, KPMG, Lovable, UiPath and Fin are already certifying against AIUC-1. These are the companies selling agents into banks, hospitals, law firms and governments, the buyers Kvist says stopped asking whether models are smart enough and started asking whether anyone can guarantee how they behave. Certification is becoming the price of admission to serious enterprise deals.
The angle worth sitting with: the market is building safety infrastructure on its own, and it is moving faster than legislation. Dattani's analogy is Underwriters Laboratories, the testing body insurers funded when electricity kept burning houses down. The UL mark still sits on most light bulbs in America. AIUC wants the same for agents: standards, testing and insurance, written by buyers, sold as a product. Anthropic chief executive Dario Amodei himself floated embedded independent evaluators in his recent pacing essay and named METR, Dattani's former organization, as a candidate. The industry is converging on independent verification from several directions at once.
For anyone buying or building AI agents, the message is practical. The enterprises writing the biggest checks are writing their own safety standard first and asking vendors to meet it. Expect procurement checklists to start naming agent certifications the way they name SOC 2 today. Builders who certify early will find enterprise doors opening faster.
Sources
New to crypto? Read the crypto glossary, browse frequent questions, read our story, or explore the story archive.