Probably Raises $9 Million for AI Error-Checking System Targeting 99.99% Accuracy
Updated
Updated · TechCrunch · Jun 16
Probably Raises $9 Million for AI Error-Checking System Targeting 99.99% Accuracy
1 articles · Updated · TechCrunch · Jun 16
Summary
$9 million in seed funding from Andreessen Horowitz will help Probably expand a system designed to stop LLM hallucinations and factual errors before they reach users.
Probably’s first product applies that approach to data science, pairing each answer with citations and an audit trail while a deterministic validator rejects outputs that do not match the underlying dataset.
Founder Peter Elias said the company trains the LLM against that validator, letting the tool run on models “four classes weaker” than frontier systems and on local hardware, cutting token costs.
The pitch comes as AI users reassess budgets and as Probably argues the same precision-focused engine could extend beyond data science into accounting, medical services and other error-sensitive work.