Artificial Intelligence Underwriting Company (AIUC), founded by Rune Kvist and Rajiv Dattani, has introduced a third‑party audit and certification offering aimed at assessing the safety of AI agents deployed inside enterprises. Kvist previously worked at Anthropic, and Rajiv Dattani is the former COO of the AI safety research organization METR.
AIUC’s approach adapts familiar cybersecurity audit models to the specific risks posed by AI agents. Drawing inspiration from the widely used SOC 2 standard, the startup developed a new standard called AIUC‑1 and a testing service to validate agents against it.
Funding and customers
On Tuesday, AIUC announced a $40 million Series A round led by Ribbit Capital, with participation from First Harmonic. The company had earlier raised $15 million in a seed round from Nat Friedman’s NFDG fund, along with Emergence, Terrain, and Ben Mann (co‑founder of Anthropic), bringing total funding to $55 million.
AIUC names Cursor, Lovable, Harvey, and ElevenLabs among its customers.
How the audits work
To build the AIUC‑1 standard, AIUC convened a consortium of roughly 250 security and risk leaders — the buyers and operators of agents — and solicited monthly feedback on what questions and assurances they would expect when purchasing agents. That input shaped the test criteria.
The company then runs an agent through a suite of about 5,000 tests that simulate scenarios including jailbreaks, hallucinations, and data leaks. The results are compiled into an approximately 100‑page report that details where an agent behaves safely and reliably, and where there are concerns.
AIUC uses AI agents to execute the tests and to help analyze the data, but humans verify and sign off on the final audit report.
Industry context
AIUC’s model resembles other proposals and efforts to introduce independent evaluation of AI behavior. Rajiv Dattani’s former employer, METR — where he served as COO from 2024 to 2025 and remains a board member — also conducts testing for frontier labs, though until recently its work focused mainly on performance (whether agents can reliably complete tasks).
Anthropic CEO Dario Amodei has publicly called for pacing frontier development and suggested embedded third‑party evaluators as one mechanism for verification, naming organizations such as METR as possibilities.
AIUC does not propose embedding evaluators within customer systems; instead, it offers an independent assessment that tells buyers where an agent can be trusted and where they should be cautious before making a procurement decision.
Why this matters
As AI models grow more capable, enterprises increasingly decline to deploy systems not because models are "not smart enough" but because they cannot currently guarantee what systems will or will not do relative to contractual or regulatory commitments. AIUC’s service aims to fill that gap by providing independent, test‑based evidence of an agent’s safety profile — a potentially important tool for organizations in banking, healthcare, government, and defense that need to evaluate and manage AI risk before adoption.
