Skip to main content
DraftCounsel

Transparency Report

What actually ran

Every figure on this page is computed from a real record — the published evaluation scorecard, the metadata-only request ledger (counts, never dollar figures; no document content; no personal data), the append-only adjudication trail (aggregate shares only), and the permission matrix the code actually enforces. Nothing here is authored copy. Companions: DraftBench · Engine status · Coverage · Security & Governance.

Published evaluation results

The committed DraftBench scorecard (run 8/19/2026) — skips are honest gaps awaiting credentials, never counted as passes.

240/245 passed· 0 failed · 5 skipped

  • 240Passed
  • 0Failed
  • 5Skipped (honest gaps)
  • clauses (deterministic assembly fidelity)44
  • cite-check (format + existence honesty)12 2○
  • grounding (cited-or-flagged validator)10
  • red-team (adversarial guardrail checks)13 3○
  • gateway13
  • drafting-gates10
  • matter-os (deterministic extraction)12
  • court-packet (readiness guidance honesty)6
  • attorney network & paralegal mode (finalization-never)14
  • small firm console (deterministic firm playbook)12
  • argument war room (framing invariant + grounding)13
  • source-locked assistant (boundary routing invariant)11
  • upl capability engine (fail-closed matrix + ledger)58
  • accessibility program (census law + token truths)12

Model-request ledger (counts only)

18

Requests (all time)

18

Last 30 days

16.7%

Error rate (all time)

3

Errors (all time)

Grounding verdicts: n/a 4 — answers the validator could not tie to the document or a cited source are flagged, and the flags are counted here.

Error kinds: timeout 3

Attorney adjudication of AI findings

No adjudications recorded yet. Every decision an attorney records on the workstation lands on an append-only trail; the aggregate shares will publish here — including the share where the attorney did not accept the AI's proposal as written.

Complaints, refunds, corrections & accessibility

Every case opened on the support page or the accessibility page is counted here — including the ones we haven't resolved. Counts only; never content or identities.

1

Cases opened

1 (1 resolved)

accessibility

Resolution share: 100.0%.

The verbatim law corpus

Sections of primary law held VERBATIM with source provenance and a normalized-text fingerprint — the store engines cite from instead of remembering. Counts are derived from the corpus table at render, never hand-typed.

850

Florida sections (verbatim)

1311

Texas sections (verbatim)

66

Federal sections (verbatim)

3

Recorded releases

append-only ledger

Store discipline

The permission matrix, as enforced

This table is generated by calling the same capability() function the routes enforce — the page cannot claim a posture the code doesn't implement. L5 is disabled for everyone: the platform never files, serves, negotiates, or appears.

Output classConsumer (LegalDraft)Attorney desk (DraftCounsel)
L0 · Published informationEnabledEnabled — attorney adjudicates
L1 · Clerical processingEnabledEnabled — attorney adjudicates
L2 · User-directed draftingEnabled — only from your explicit selectionsEnabled — attorney adjudicates
L3 · Personalized analysisEnabled — only for a state you selectEnabled — attorney adjudicates
L4 · Recommendation / strategyUnranked alternatives only — you choose; software-selected recommendations only under an express state safe harbor (e.g. TX), labeled as automated suggestionsEnabled — attorney adjudicates
L5 · RepresentationNever — for anyoneNever — for anyone

13 surfaces are classified in the registry; unregistered surfaces and unknown states fail closed. Hard gates currently enforced: 14 of 14 registered (jurisdiction, coverage, claim-state-exact, attestation, access, attorney-finalization, citation-label, execution, upl, required-facts, conflict, currentness, court-form, upl-capability).

Why this page exists: regulators and courts are actively deciding how AI legal tools should be supervised, and the reform frameworks converge on disclosure — publish what the tool does, how often it errs, and who exercises judgment. DraftCounsel publishes those numbers from its own records rather than asking to be taken on faith. This report is factual disclosure, not a certification, and DraftCounsel is not a law firm.