One day after an Anthropic researcher publicly resigned over fears that AI could kill us all before 2030, two founders sat down to pitch a different kind of response. Not a moratorium. Not a manifesto. A testing framework. According to TechCrunch, Rune Kvist and Rajiv Dattani, brothers-in-law with unusually relevant backgrounds, have built a startup called Artificial Intelligence Underwriting Company (AIUC) that wants to bring structured safety auditing to enterprise AI agents.
Kvist is an early Anthropic employee. Dattani is the former COO of METR, the AI safety research organization that has done independent evaluations for OpenAI and others. Together, they are arguing that the core problem with enterprise AI adoption isn’t model capability. It’s accountability. “Banks, hospitals, governments and militaries no longer decline to deploy AI because a model isn’t smart enough,” Kvist said. “They decline because they’ve made commitments to their own customers about what a system will and won’t do, and nobody can currently guarantee that.”
AIUC’s answer to that gap is a certification standard called AIUC-1, modeled explicitly on SOC 2. The analogy is deliberate. SOC 2 gave enterprises a shared language for evaluating cloud security vendors. AIUC is betting that AI agents need the same thing. To build the standard, they assembled a consortium of roughly 250 security and risk leaders who are actually buying agents at scale, then asked them what they’d want to see before signing a contract. That input shaped a test suite of around 5,000 scenarios covering jailbreaks, hallucinations, and data leaks. The output is a report of about 100 pages that maps where an agent behaves reliably and where it doesn’t. AI agents run the tests, AI analyzes the results, and humans verify the final audit.
The company counts Cursor, Lovable, Harvey, and ElevenLabs among its customers, which is a credible early list. On the funding side, AIUC closed a $40 million Series A led by Ribbit Capital, with First Harmonic also participating. That follows a $15 million seed from Nat Friedman’s NFDG, Emergence, Terrain, and Anthropic co-founder Ben Mann, bringing total raised to $55 million. Ribbit is primarily a fintech fund, which makes sense given that financial services is one of the sectors most visibly blocked by AI compliance concerns.
The comparison to METR is worth addressing directly. METR does similar evaluations, but its focus has historically been on whether agents can complete tasks reliably, not on enterprise safety certification as a commercial product. Anthropic’s Dario Amodei has also floated the idea of requiring frontier labs to use embedded third-party evaluators, naming METR as a candidate. AIUC isn’t proposing to embed itself anywhere. But the underlying logic is the same: independent verification matters.
Whether AIUC can actually own this category depends on a few things. Standards bodies, regulators, and hyperscalers all have reasons to want this space for themselves. The EU AI Act is already pushing in this direction from the regulatory side. And the value of AIUC-1 as a standard only compounds if adoption is broad enough to matter. Still, the founders have the right networks, the right timing, and a product that maps to a real procurement blocker. That’s a better starting position than most.
