Early Anthropic hire, former METR COO have found a way to rein in rogue AI agents

A day after Anthropic researcher Jacob Coxon’s resignation over extinction-risk concerns, TechCrunch sat with AIUC founders Rune Kvist (early Anthropic) and Rajiv Dattani (former METR COO). Their Artificial Intelligence Underwriting Company wants to bring third-party audit and certification to enterprise agents.

AIUC announced a $40 million Series A led by Ribbit Capital, plus a prior $15 million seed that included Nat Friedman’s NFDG and Anthropic co-founder Ben Mann — $55 million total. Customers named include Cursor, Lovable, Harvey, and ElevenLabs.

We’re treating this as the insurance-and-audit layer trying to catch up with agent deployment anxiety.

We’re glad someone is productizing buyer questions instead of another vibe-y safety PDF. Kvist’s line lands: banks, hospitals, governments, and militaries are not waiting for smarter models — they need guarantees about what a system will and won’t do.

AIUC-1 mirrors SOC 2 logic. A consortium of about 250 security and risk leaders shapes the asks. Agents face roughly 5,000 tests covering jailbreaks, hallucinations, and data leaks; AI helps run and analyze tests, humans verify the final audit; buyers get a ~100-page map of pass and fail. That is closer to procurement reality than lab eval theater.

Dattani’s METR background matters — METR helped OpenAI investigate Hugging Face — but AIUC is aimed at enterprise agents, not embedding inside frontier campuses the way Amodei’s auditor pitch imagines. Useful independent assessment can unlock deployments that are stuck. The downside if this scales poorly is checkbox certification that lags new failure modes, or a cottage industry of passing the test suite. I’m watching whether AIUC-1 becomes a default RFP requirement and how often reports actually kill deals.

Amodei’s pacing essay named embedded third-party evaluators such as METR. AIUC is adjacent in spirit but positioned as pre-purchase assessment rather than in-lab embeds.