Corrections & Error Reports

Email legal@familylawgps.aiwith: the page or tool, what the output said, what you believe is wrong, and — if a citation is involved — the citation's hash shown beside it (every retrieved citation carries one). Nothing else is required; screenshots help.

Reports are checked against the hash-pinned corpus and the deterministic engines — the same verification everything on the platform runs on. A confirmed law-text issue triggers a re-capture from the official source, and the version ledger records the change with the old and new hashes; a confirmed template or engine defect is fixed in code, where the public FamilyBench suites gate every push. Confirmed corrections change the platform for everyone, not just the reporter.

We do not promise response times we have not engineered — reports are read and confirmed issues are fixed in the order their severity demands, and the git history and version ledgers are the public record that it happened.

Work-product decisions stay yours and travel with your documents: the Decision-Authority Ledger prints with every assembled draft (who selected the task, whose facts, what created the text, what you decided), and tool entries persist in your browser unless you save them to an account. The platform does not train models on your documents.

The Authority Engine shows the live release gate, corpus load map, and the per-state capability matrix; FamilyBench publishes the eval scorecards with a reproducibility bundle; and the Provenance Seal re-verifies any export's citations against the live corpus, no account required.