Probably Raises $9M to Quantify AI Confidence—a Regulatory Requirement Disguised as Product

Probably closed a $9 million seed round on June 16, 2026, to build tooling that surfaces confidence signals alongside LLM outputs. Current models produce answers with uniform fluency regardless of accuracy. Probably's approach quantifies reliability at inference—critical for regulated industries (legal, healthcare, finance) where compliance teams now demand auditability and error rates. Whether confidence quantification remains standalone infrastructure or gets absorbed into model APIs remains open.