Show HN: Factlabel: catches AI agents lying about the data they're reporting on

Category: ai-ml

Tags: ai-audit, fact-checking, llm-verification, data-provenance, trust-layer

Score: 7.5/10 (Innovation: 8, Technical: 7, Documentation: 8, Utility: 7)

Factlabel is a Python-based audit layer that sits between AI agents and readers, verifying every claim, figure, and attribution in AI-generated content against source data. It combines deterministic code checks with a parallel decision model (Jev) to produce typed pass/review/block verdicts, annotated findings, and a trust badge. It is interesting because it treats LLM fact-checking as a structured decision problem rather than open-ended prose, making verification cheap, fast, and resistant to persuasion.

Target audience: backend developers, data engineers, AI engineers, product managers building agent workflows

Repository: https://github.com/generallymatthew/factlabel · Python · Apache-2.0 · 1 stars

View on Hacker News