AIUC raises $40M to put a trust mark on enterprise AI agents
Agent safety is turning into an industry with a business model. Today's money went to the auditors and to the plumbing under the agents themselves.
AIUC closed a $40 million Series A led by Ribbit Capital, with First Harmonic participating, to sell enterprises an independent safety verdict on the agents they buy. The company — formal name Artificial Intelligence Underwriting Company — built AIUC-1, a standard and testing service modeled on the SOC 2 playbook, and its pitch is blunt: enterprises do not stall deployments because models are too dumb, they stall them because nobody can promise the system will behave. AIUC runs agents through roughly 5,000 risk-and-attack combinations covering jailbreaks, hallucinations and data leaks, tailored per industry, and returns a report of about 100 pages showing where the agent holds up and where it does not. Agents are re-audited and recertified every quarter. Cursor, ElevenLabs, Harvey, KPMG, Lovable, UiPath and Fin already certify against the standard, and about 250 security and risk leaders from Fortune 1000 companies shape what the tests ask.
The founders are the reason the round got done. Rune Kvist was Anthropic's first product hire; Rajiv Dattani was a partner in McKinsey's insurance practice and chief operating officer at METR, the AI safety evaluator. Dattani's framing is the interesting part — when electricity was burning down houses, the insurers paying the claims funded Underwriters Laboratories to test and certify products, and the UL mark is still on most light bulbs in America. AIUC wants the same trio of standards, testing and insurance, pointed at models instead of appliances. The $40 million is meant to push that from agent audits toward frontier-model oversight.
Two caveats worth holding onto. The prior round was $15 million from Nat Friedman's fund NFDG, Emergence, Terrain and Anthropic co-founder Ben Mann, so total funding is $55 million — respectable, not enormous, for a company that wants to sit between every enterprise and every agent vendor. And a certification is only as strong as the buyers who demand it: the standard matters the moment procurement teams write "AIUC-1 certified" into contracts, and not before. Today's list of certifying companies suggests that is already starting to happen — see how the governance layer is consolidating elsewhere, as with StackGen putting one governance plane under every production agent.
Keewano launched KeewanoDB, an event-oriented database built for agents rather than for people writing SQL, and pulled in $12 million from Hetz Ventures and a16z Speedrun. The Tel Aviv company's argument: relational databases and warehouses were designed for structured queries by humans, so an agent asking why a customer churned has to flatten history into tables or wait on an ETL pipeline first. KeewanoDB groups events around an entity and keeps them in the order they happened, so an agent can walk a history directly. The company says each event takes about four bytes, runs on CPU with vectorized instructions rather than GPUs, and can query roughly 250 million events in under half a second — performance the company reports but has not had independently verified.
The interesting bit is where the reasoning runs. An in-database engine executes Lua scripts that agents send over the Model Context Protocol, so filtering happens next to the data and the model receives a narrow window of context instead of the whole event history — Keewano says customers pulling from Snowflake or BigQuery report token savings around 84 percent. It charges by active entities rather than events, and supports Parquet, Iceberg and Kafka ingestion. It coexists with the warehouse rather than replacing it, which is the only honest way to sell this into a company that just spent three years building one.
Evvy closed $40 million to bring AI-driven precision diagnostics to women's health, and the plan starts with fertility. The round is a Series B led by Catalio Capital Management, with the U.S. Fertility Innovation Fund, Rethink Impact, Muse Capital and Alumni Ventures joining and existing backers LabCorp Venture Fund, General Catalyst and Left Lane Capital following on. Evvy's CEO Priyanka Jain puts the problem in numbers: the biological signals specific to female bodies were largely unstudied because clinical trials generally excluded women until 1993, leaving roughly 30 years of data, and the cost shows up as misdiagnosis rates above 50 percent for some conditions and recurrence rates above 60 percent for common symptoms.
EvvyAI analyzes a swab, matches the patient to a personalized treatment plan, and keeps learning from each new test. The company says clinicians using it diagnose more than 90 percent of symptomatic patients successfully and cut recurrence of bacterial vaginosis by 50 percent, and that its microbiome work has identified six distinct BV subtypes — patients who looked identical under conventional criteria but responded to different treatments. It has served more than 100,000 patients through about 3,000 practitioners, and its dataset has been cited in 13 peer-reviewed publications. That is the real asset here: not the model, the labeled biology nobody else collected.
Should a certification badge, a quarterly audit, or a customer's own procurement team be the thing that decides whether an AI agent is safe enough to deploy? Tell us in the comments.
