Pramaana Labs, a startup building a way to verify the output of artificial intelligence, has raised a $27 million seed round led by Khosla Ventures, according to TechCrunch.
The company's pitch centers on a technique borrowed from mathematics and computer science called formal verification. According to TechCrunch's Russell Brandom, Pramaana uses the LEAN programming language to build what it describes as a deterministic verification layer that sits on top of large language models — the kind of AI systems that power today's chatbots and assistants. In plain terms, the idea is to add a checking step that can confirm whether an AI's answer is actually correct, rather than simply plausible-sounding.
Pramaana plans to aim this technology at high-stakes fields where mistakes carry real consequences. TechCrunch reports the company will focus on sensitive verticals like law, drug discovery, and tax preparation — areas where errors can be costly and reliability is at a premium.
The funding arrives against a familiar backdrop. As Brandom notes, enterprises have struggled to turn their AI pilot programs into functional parts of their business, and reliability has become a central obstacle. Large language models are known to produce confident but wrong answers, which makes them hard to trust in regulated or safety-critical work.
Why it matters: if a verification layer can reliably flag when an AI is wrong, it could help move the technology out of experimental pilots and into the high-stakes professional settings where it has so far been too risky to deploy.