AIUC

AIUC (the Artificial Intelligence Underwriting Company) is “an AI agent standards & insurance company based in San Francisco” building what it calls “confidence infrastructure for AI” to accelerate secure enterprise adoption of AI agents. AIUC’s framing is that many organizations feel competitive pressure to adopt AI agents, but hesitate because failures can cause “data leakages, jailbreaks, hallucinations, and reputational harm.” In response, AIUC combines a technical certification/audit pathway with financial risk transfer: “Certificates and insurance create the “confidence infrastructure” enterprises need to adopt AI agents.”

On the standards side, AIUC offers AIUC-1, described as “the world’s first standard for AI agents,” covering “data & privacy, security, safety, reliability, accountability and societal risks.” AIUC-1 is designed to be operational and test-driven; for example, its safety focus includes “rigorous 3rd-party testing, monitoring, and safeguards including human-review of flagged outputs.” The AIUC-1 certificate is also positioned as a continuously maintained compliance artifact rather than a one-time stamp: “Like ISO 27001, FedRAMP, and CSA STAR, AIUC-1 requires ongoing technical testing and compliance. It must be renewed annually to remain current.” In its overview, AIUC specifies that auditors evaluate compliance using “upfront technical testing and review of operational controls (conducted annually), and ongoing technical testing (conducted at least quarterly to keep up with ongoing changes to AI risk & mitigation techniques).”

On the insurance side, AIUC provides policies meant to protect enterprise customers against AI-agent-related losses: “AIUC helps AI companies protect their customers financially in case their AI agent fails.” It states that “Our insurance policies are backed by the world’s most trusted insurers, unlocked by our underwriting,” and that it can “cover AI-specific risks up to $50m of losses.” This pairing is explicitly embedded in AIUC’s go-to-market logic for enterprise adoption: enterprises “can’t do so with confidence,” AI companies “certify against AIUC-1,” then “buy insurance to fully protect their enterprise customers,” resulting in an enterprise that “can now adopt with confidence.”

AIUC also highlights enterprise-facing assessments and delivery artifacts, including rapid technical testing and an audit/certificate workflow: “Technical tests: Get eval test results within 24 hours,” followed by “Gap analysis” and then an “AIUC-1 certificate.” It showcases early customer/partner examples. Intercom, for instance, “achieved AIUC-1 certification of their AI agent, Fin, to validate technical safeguards against data leakage, jailbreaks, and hallucinations through independent adversarial testing,” and the audit is described as spanning “50+ controls” with “Quarterly adversarial testing across 1,000+ scenarios designed by security researchers.” AIUC also cites additional examples of certification being used to raise assurance for enterprise buyers, including Ada certifying “against the AIUC-1 framework” via “the most rigorous testing standards --- far exceeding the typical enterprise requirements,” and Alpharun using AIUC-1 certification to support compliance needs and help a large customer gain confidence “after reviewing the audit and legal analysis.”

The organization’s team narrative emphasizes a blend of AI risk and insurance experience and positions insurance as a lever for safe progress: “Our team brings together experience across AI risk and insurance — having spent time at Anthropic, McKinsey insurance, METR, and Center for AI Safety,” and it has published “Underwriting Superintelligence” as “a research paper exploring how insurance can unlock secure AI progress.” AIUC reports it “raised a $15M seed round led by Nat Friedman,” with additional named investors.