Mentioned:ValsAndreessen Horowitz$40M · Series A
Vals raised $40 million in Series A funding led by Andreessen Horowitz to expand its AI model evaluation platform. The startup, which now employs 25 people after starting the year with eight, recently launched a program providing benchmarking services to federal agencies.
What's happening
- Vals closed a $40 million Series A led by Andreessen Horowitz following rapid growth that tripled its staff to 25 people in nine months.
- The company launched a new program providing model evaluations to federal agencies, expanding from commercial clients to the defense industrial base.
- Vals now evaluates AI capabilities across cybersecurity, biosecurity, law of armed conflict, and recursive self-improvement domains beyond traditional benchmarks.
- Company revenue grew eight times year over year, driven by a pricing model where AI developers pay for independent testing of their models.
Why it matters
- Federal acquisition of AI models increasingly requires validated performance metrics; Vals' federal program provides independent verification that shapes procurement decisions.
- Defense applications in cybersecurity and law of armed conflict demand benchmarks designed to detect misuse and compliance failures, filling a gap in legacy evaluation systems.
- As AI becomes embedded in military systems, third-party benchmarking establishes trust and accountability for capability claims, reducing procurement risk.
- Vals' refusal to publish test materials prevents model developers from training specifically to pass benchmarks, ensuring evaluations measure real-world performance.
Going deeper
- Vals compares its revenue model to the College Board SAT system, where test-takers pay for independent verification of competency rather than self-reported results.
- The startup measures not only positive model capabilities but also negative outcomes, simulating how unrestricted deployment would affect sensitive domains like mental health and biosecurity.
- Vals was founded in 2024 by 25-year-old Rayan Krishnan, who previously interned at Palantir and worked at Microsoft and Stanford's AI lab.
Financial impact
- Series A capital of $40 million will fund hiring plans to add 10 to 15 employees and relocate to larger offices, expected to accelerate in fourth quarter 2026 and into 2027.
- Revenue grew eight times year over year from prior baseline; timing and absolute revenue figures not disclosed, but growth rate indicates path to profitability before capital depletion.
- Federal program revenue stream and commercial client base both contribute to top-line growth, though breakdown between segments not disclosed.
The intrigue
- Federal agencies and AI vendors increasingly need benchmarks certified as tamper-proof, creating a market dynamic where companies pay for results they may not want, aligning incentives for honest evaluation.
The fine print
- Vals does not publicly disclose its specific test materials to prevent reverse-engineering, limiting outside verification of benchmark rigor and methodology.
- Federal program details including customer agencies, contract scope, and performance metrics not disclosed.
Source: TechCrunch · Sep 19, 2026 · Drafted with Claude, reviewed by Industrial Base Alpha.