AIUC Raises $40 Million to Audit Rogue Enterprise AI Agents
AIUC raised $40 million to certify enterprise AI agents through 5,000 safety tests, giving buyers independent checks on jailbreaks, leaks and hallucinations.
Summary
On Tuesday, September 15, 2026, Artificial Intelligence Underwriting Company announced a $40 million Series A led by Ribbit Capital, with First Harmonic participating. Founders and brothers-in-law Rune Kvist, an early Anthropic employee, and Rajiv Dattani, METR COO from 2024 to 2025 and a current board member, previously raised a $15 million seed from Nat Friedman’s NFDG, Emergence, Terrain, Anthropic co-founder Ben Mann and others, bringing total funding to $55 million. Customers include Cursor, Lovable, Harvey and ElevenLabs. The announcement followed Anthropic researcher Jacob Coxon’s resignation over fears AI could kill humanity by the decade’s end.
AIUC offers third-party audits and certification through AIUC-1, a standard modeled on SOC 2 and shaped by monthly feedback from about 250 security and risk leaders who buy agents. AI agents run and analyze roughly 5,000 tests covering jailbreaks, hallucinations and data leaks, producing a roughly 100-page report that humans verify before buyers assess passes and concerns. Kvist says banks, hospitals, governments and militaries increasingly reject AI because vendors cannot guarantee behavior promised to customers. METR performs similar frontier-lab evaluations, historically focused on reliable task completion, and helped OpenAI investigate its Hugging Face incident. Anthropic CEO Dario Amodei has urged slower frontier development and proposed embedded third-party evaluators, including METR, as bad-behavior incidents rise. AIUC provides independent assessments but does not embed at customer sites.
Positives
- The $40 million Ribbit Capital-led Series A brings AIUC’s total funding to $55 million.
- About 250 security and risk leaders provide monthly input shaping the AIUC-1 certification standard.
- Roughly 5,000 tests examine jailbreaks, hallucinations and data leaks before humans verify each final audit.
- Cursor, Lovable, Harvey and ElevenLabs already use AIUC’s independent agent assessments.
- The roughly 100-page reports show enterprise buyers where agents pass and where concerns remain.
Risks & concerns
- Jacob Coxon quit Anthropic over fears AI could kill humanity by the decade’s end.
- Banks, hospitals, governments and militaries lack guarantees that agents will honor commitments made to customers.
- Dario Amodei cited rising bad-behavior incidents while calling for slower frontier AI development and independent evaluation.
- AIUC does not embed evaluators at customer sites, unlike the oversight model Amodei proposed for frontier labs.
- Jailbreaks, hallucinations and data leaks remain significant enough to require thousands of pre-purchase tests.