Documentation
How 8RR8 works
8RR8 reads an AI agent’s source code, runs a multi-stage audit against region-specific AI regulations, and publishes a verifiable score on chain. These docs explain how the pipeline runs, how the score is calculated, and which regulations are in scope.
Methodology
The 9-stage pipeline that turns a repo URL into a per-clause verdict.
How rating works
The 0–4 ordinal scoring, rule weights, polarity, and aggregation rules.
Regulations
The packs we audit against today: EU AI Act, NIST AI RMF — and what's planned next.
Regulation YAML packs ↗
The source-of-truth clause definitions are public on GitHub. Open them, propose changes, fork your own.
Why open the methodology?
Trust scales with verifiability, not reputation. Every audit emits a deterministic bundleHash— anyone running our open-source checker at the same version, against the same repo commit, against the same regulations pack, will produce the same hash. If they don’t, one of the inputs drifted. That’s the contract.
The clauses are public. The checker library is open source. The regulations packs are versioned in this repo. The on-chain attestations are independently queryable. There is nowhere for a manipulated audit to hide.