Two former AI safety leaders, Rune Kvist and Rajiv Dattani, are betting that the next phase of enterprise AI will require a new layer of trust. Their startup, Artificial Intelligence Underwriting Company (AIUC), is designed to help companies evaluate whether AI agents behave reliably before they are deployed at scale.
Kvist, an early Anthropic employee, and Dattani, formerly COO of the AI safety research group METR, say the challenge is no longer just building smarter systems. As AI becomes more capable, they argue, it also becomes harder to govern inside real-world workflows. AIUC aims to address that gap with independent testing and certification for AI agents.
How the system works
The company has created AIUC-1, a standard inspired by the cybersecurity framework SOC 2. It also runs agents through thousands of tests covering jailbreak attempts, hallucinations, and data leakage risks. The output is a detailed report showing where a model performs well and where extra caution is needed.
AIUC says its process is shaped by feedback from a consortium of about 250 security and risk leaders, reflecting what enterprise buyers want to know before adopting AI tools. The startup counts companies such as Cursor, Lovable, Harvey, and ElevenLabs among its customers.
On Tuesday, AIUC announced a $40 million Series A led by Ribbit Capital, with participation from First Harmonic. The round follows an earlier $15 million seed investment, bringing total funding to $55 million.
For enterprises, the appeal is clear: a neutral assessment that helps define what an AI system can do, where it should be limited, and how confidently it can be trusted. If this model gains traction, it could become a standard part of how organizations adopt AI in the years ahead.
