Pramaana Labs Raises $27 Mn to Make AI Prove Its Answers
Bengaluru-based Pramaana Labs raised $27 Mn in seed funding led by Khosla Ventures to build a verification layer for AI in regulated industries. The company says its system turns questions into formal statements, runs a proof engine and refuses to answer when a proof cannot be established. The financing points to a tougher enterprise AI test: whether models can support tax, law, finance and healthcare work where hallucinations carry legal risk.
Pramaana Labs has raised $27 Mn to pursue a stricter version of the enterprise AI promise: answers that can be verified, not just generated.
Funding Goes To A Verification Layer
The Bengaluru startup said the seed round was led by Khosla Ventures, with Accel, BoldCap, Nexus Venture Partners, Premji Invest and Unbound also participating.
The company is building what it describes as a verification layer for AI systems used in tax, law, finance and healthcare.
Pramaana plans to use the capital to train formalisation and prover models, hire more AI research talent and deepen its expertise in regulated sectors.
The spending targets an infrastructure problem inside enterprise AI: how to turn rules, statutes and domain logic into outputs a machine can verify.
Ranjan Rajagopalan, Krishnan Raghavan and Sanjay Ganapathy Subramaniam, all IIT Madras alumni, founded the Bengaluru-registered company in September 2025.
The pitch targets high-stakes settings where an AI hallucination can become a compliance, legal or medical risk rather than a minor product flaw.
The Product Test Is Proof, Not Fluency
Pramaana’s engine is designed to translate a user query into formal statements, execute a proof engine and return a machine-checkable proof of correctness.
If the system cannot establish the proof, it is designed to identify the breached rule or refuse to answer.
That design sets Pramaana apart from AI tools that rely mainly on fluent language output.
Rajagopalan wrote that tax, law, finance and healthcare depend on certainty, while probabilistic AI cannot provide that certainty.
He also said the company wants every output to ship with mathematical proof of correctness.
Vinod Khosla described auto-formalisation as a missing AI capability in a funding announcement video.
Pramaana’s technical direction has been compared with the LEAN programming language, which is used to verify mathematical proofs and can represent complex domain knowledge in machine-checkable form.
India’s AI Funding Is Moving Toward Applications
The size of the seed round is part of the story.
Five seed-stage startup funding rounds in the previous week totaled $22 Mn, making Pramaana’s $27 Mn raise large for its stage.
The deal also came close to another financing event involving Sarvam, another Khosla-backed India AI company.
Sarvam’s round was a $300 Mn Series B tied to a $1.5 Bn valuation.
Those numbers do not prove that Indian AI infrastructure has solved enterprise adoption, but they show capital moving into companies that promise more reliable production use.
Inc42 Datalabs put Indian AI startup funding at $253 Mn in Q1 2026, up around 73% from a year earlier.
More than 86% of funding raised by Indian AI companies since 2020 has gone to application-layer businesses, which makes Pramaana’s verification focus notable because it sells a trust layer for applications rather than a consumer-facing assistant.
What Buyers Still Need To See
Pramaana has disclosed investors, founders, target sectors and the verification method it wants to build.
The public record still lacks named enterprise customers, product availability, pricing or deployment results.
Those gaps define the next test.
In tax, law, finance and healthcare, buyers will need proof that formalisation and prover models can handle real domain rules, not just funding-demo examples.
The $27 Mn round gives Pramaana room to hire researchers and train models; customer evidence will decide whether provable AI moves from a technical claim to an operating system for regulated work.
