One day after Anthropic researcher Jacob Coxon resigned over fears that AI might end human life within the decade, I sat down with Rune Kvist and Rajiv Dattani, founders and brothers-in-law. They believe they've found a way to protect us—or at minimum, to keep AI agents from turning rogue within corporate environments.

"AI is getting smarter at an increasingly rapid rate. The surprising thing about AI is that it becomes harder to adopt and harder to control as AI gets smarter, not easier," said Kvist, who joined Anthropic early on and is married to Dattani's sister. Dattani previously served as COO of METR, an AI safety research organization.

Together they've launched the Artificial Intelligence Underwriting Company (AIUC), aiming to bring AI safety measures to enterprises and the companies developing AI models and agents. The startup lists Cursor, Lovable, Harvey, and ElevenLabs among its customers.

On Tuesday, AIUC revealed a $40 million Series A round led by Ribbit Capital, with First Harmonic also participating. The company had previously secured $15 million in seed funding from Nat Friedman through his NFDG fund, alongside Emergence, Terrain, and Anthropic co-founder Ben Mann, among others. This brings total funding to $55 million.

What drew this impressive group of investors was AIUC's approach: applying a well-established cybersecurity framework to a fresh set of AI risks. The company has created a third-party audit and certification process for AI agents.

"Banks, hospitals, governments and militaries no longer decline to deploy AI because a model isn't smart enough," Kvist explained. "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."

Taking inspiration from SOC 2, a widely adopted cybersecurity standard, AIUC developed its own standard called AIUC-1, along with a testing service to validate agents against it. To create the standard, AIUC brought together a consortium of roughly 250 security and risk leaders—the people who purchase agents.

"These are the people who we meet with on a monthly basis, and the question we ask them is: When you're buying agents from someone, what would you look for. What are the questions you'd want to ask, and what would you want to see addressed." That input directly informs the tests.

The startup then puts an agent through approximately 5,000 tests to observe its behavior in scenarios involving jailbreaks, hallucinations, and data leaks. The outcome is a report of roughly 100 pages detailing where an agent performs safely and reliably—and where it falls short. Notably, AIUC relies on AI agents to conduct the tests and AI to analyze the results, though humans verify the final audit, according to Kvist.

If this sounds somewhat familiar, that's because it is. Dattani's former employer METR—where he served as COO from 2024 to 2025 and still sits on the board—conducts similar testing for frontier labs, though its work until recently centered mainly on performance, specifically whether agents can reliably complete tasks. METR was among the independent research organizations OpenAI enlisted to investigate its Hugging Face incident.

Anthropic CEO Dario Amodei has also recently urged the AI industry to slow frontier development, pointing to a sharp rise in bad-behavior incidents. In his post, Amodei suggested requiring frontier labs to use embedded third-party evaluators to observe and verify safety, mentioning METR as one option.

While AIUC doesn't propose embedding itself at customer sites, the underlying concept is comparable: providing enterprises with an independent assessment of how safe their AI agents truly are.

"Here's where it passes and where you can trust it. And here's where there's concerns. You should be aware of those references before you make the decision to buy," Dattani said.