AIUC Raises $40 Million to Certify AI Agents Against Safety Risks

AIUC Secures $40 Million to Scale AI Safety Certification

Artificial Intelligence Underwriting Company (AIUC) has raised $40 million in Series A funding2

2

led by Ribbit Capital, with participation from First Harmonic1

1

. Combined with its $15 million seed round from Nat Friedman through NFDG, Emergence, Terrain, and Anthropic co-founder Ben Mann, AIUC has now raised $55 million in total1

1

. The startup addresses a critical bottleneck in AI adoption: enterprises declining deployment not because models lack capability, but because they cannot guarantee what systems will and won't do1

1

.

Founders Bring Deep AI Safety Expertise

Co-founders Rune Kvist and Rajiv Dattani bring complementary backgrounds to tackle rogue AI agents. Kvist served as an early Anthropic employee and the company's first product hire2

2

, while Dattani worked as COO of AI safety research organization METR from 2024 to 2025 and remains a board member1

1

. Dattani also spent time as a partner in McKinsey's insurance practice2

2

. The brothers-in-law launched AIUC in July 2025 to bring AI safety to enterprises and companies building AI models and agents1

1

.

AIUC-1 Standard Tackles Enterprise Security Concerns

AIUC developed the AIUC-1 standard by assembling a consortium of about 250 security and risk leaders from Fortune 1000 companies1

1

. These buyers of agents meet monthly with AIUC to shape testing requirements based on real-world deployment needs1

1

. The AI agent certification process runs agents through about 5,000 tests covering risk and attack scenarios tailored to specific business contexts2

2

. Tests evaluate behavior involving jailbreaks, hallucinations, and data leaks1

1

. Each third-party audit produces a roughly 100-page report detailing where an agent performs safely and where concerns exist1

1

. AI agents conduct the testing and analyze data, though humans verify the final audit1

1

. Agents are audited independently and recertified every quarter as attack techniques evolve2

2

.

Major AI Companies Already Using Certification

AIUC counts Cursor developer Anysphere, Lovable Labs, Harvey, and ElevenLabs among its customers1

1

. Voice AI company ElevenLabs used its certification in February to secure first-of-its-kind insurance for AI agents, including coverage for voice agents giving customers incorrect information2

2

. KPMG, UiPath, and customer service software maker Fin (formerly Intercom) also hold the certification2

2

. Kvist notes that most enterprises maintain lists of agents approved in pilots but stalled at security reviews, making proof of security and reliability the main bottleneck2

2

.

Expansion to Frontier AI Models Addresses Growing Pressure

AIUC initially focused on applications where security reviews first blocked rollouts2

2

. The company now plans to extend its audit and insurance work to frontier AI models as similar pressure builds at that layer2

2

. This timing aligns with Anthropic CEO Dario Amodei recently calling for the AI industry to pace frontier development, citing rapid increases in bad-behavior incidents1

1

. Amodei floated requiring frontier labs to use embedded third-party evaluators to observe and verify safety, naming METR as one possibility1

1

. While AIUC doesn't propose embedding at customer sites, it offers independent assessment of AI agent safety1

1

.

Why This Matters for Enterprise AI Governance

The funding reflects investor recognition that AI adoption now faces a trust problem rather than a capability problem. Micky Malka, founder and managing partner of Ribbit Capital, said his firm spent over a decade backing financial services companies where trust counts above all else, and AI is heading down the same road while moving faster than evaluation systems2

2

. Dattani draws parallels to historical governance models: "When electricity was burning down houses, the insurers paying the bill funded Underwriters Laboratories to test and certify products. To this day, the UL mark is on most light bulbs across America. AI needs the same combination of standards, testing and insurance"2

2

. Banks, hospitals, governments and militaries no longer decline AI deployment because models aren't smart enough—they decline because they've made commitments about system behavior that nobody can currently guarantee1

1

. Watch for AIUC's expansion into frontier model auditing to influence how labs approach safety validation and whether regulators adopt similar standards for high-stakes AI deployments.