AIUC Secures $40M to Audit and Certify Enterprise AI Agents

AIUC raises $40 million for AI-agent audits

Artificial Intelligence Underwriting Company (AIUC) has announced a $40 million Series A led by Ribbit Capital, with participation from First Harmonic. The company, founded by early Anthropic employee Rune Kvist and former METR chief operating officer Rajiv Dattani, now has total funding of $55 million.

Its earlier $15 million seed round included Nat Friedman’s NFDG, Emergence, Terrain and Anthropic co-founder Ben Mann. AIUC names Cursor, Lovable, Harvey and ElevenLabs among its customers.

AIUC is building a third-party audit and certification layer for AI agents used by enterprises and companies developing models and agents. Kvist argues that adoption barriers increasingly concern whether organisations can establish what an AI system will and will not do, rather than whether the underlying model is sufficiently capable.

An AI safety standard inspired by SOC 2

The startup has developed a standard called AIUC-1, taking inspiration from the widely adopted cybersecurity framework SOC 2. AIUC assembled a consortium of roughly 250 security and risk leaders to shape the questions and assurances that buyers expect when assessing AI agents.

The resulting assessment runs an agent through approximately 5,000 tests. These examine behaviour in scenarios involving jailbreaks, hallucinations and data leaks. AIUC then produces a report of about 100 pages describing where an agent performs safely and reliably, as well as the areas in which it does not.

AI agents conduct the tests and analyse the resulting data, but humans verify the final audit. This division is intended to combine automated evaluation at scale with a human-reviewed conclusion for organisations considering deployment.

Independent evaluation becomes a procurement issue

AIUC’s approach resembles the independent evaluation work associated with METR, where Dattani served as COO from 2024 to 2025 and remains a board member. METR’s work has focused largely on whether agents can reliably complete tasks, including evaluation work connected with OpenAI’s Hugging Face incident.

The wider context also includes calls for more external scrutiny of frontier systems. Anthropic CEO Dario Amodei has proposed embedded third-party evaluators for frontier laboratories, while AIUC is instead offering an assessment that buyers can use when judging an agent supplier. The demand for such evidence sits alongside Anthropic’s growth beyond model performance, which traces Anthropic’s growth beyond the performance of its models alone.

For businesses, the practical implication is to treat agent evaluations as procurement evidence: review the specific failure modes, controls and audit findings before assigning an agent access to sensitive data or consequential workflows.