The AI evidence trail your Rule 11 and client audits will ask for.

What is the firm-wide AI governance dashboard designed to be?

The dashboard is designed to be one screen that answers a question a firm cannot currently answer quickly: across every open matter, where was AI used, what was verified before it went out the door, and what is still unresolved? Today that answer is scattered across individual certificates and attestations. Each of those records is real, signed, and independently verifiable on its own, but there is no aggregated view that rolls them up into a picture a partner or a court could scan. The governance dashboard is meant to be exactly that roll-up — a live, filterable cockpit that reads the records the platform is already generating and presents them as a coherent trail rather than a pile of individual proofs.

It is important to be precise about the layering here, because it is the whole honesty story of this page. The dashboard does not create the evidence. The evidence is created at the moment of the work: when a citation is certified before a filing, when an interaction with an approved AI tool is isolation-attested, when consent is captured. Those records are signed and sealed as they happen. The dashboard is a reader and an organizer of those records — a reporting surface. Building the reporting surface changes nothing about the strength of the underlying proof; it changes how quickly and how completely a human can see it.

• AI-assisted matters — every matter where an approved tool was used, with the certification status of each filing's citations attached.
• Isolation attestations — privilege-isolation records by matter and by approved tool, so a firm can see which interactions ran inside the attested boundary.
• Open items — citations flagged, filings held, consent not yet captured — surfaced so nothing sits unresolved out of view.
• On-demand evidence packs — a per-matter bundle a firm can hand to a court or client, each item independently verifiable rather than asserted.

What would the dashboard show at a glance?

The design intent is a top-level view that reads like an instrument panel, not a spreadsheet export. A managing partner should be able to open it and, without training, see the shape of the firm's AI exposure: how many matters touched AI, how many of those have every citation certified, how many isolation attestations are on record, and how many items are still open and waiting on someone. From there, the same view should filter down — to a single matter, a single tool, a single date range, or a single unresolved question — so that the person who needs a specific answer can reach it without asking anyone to reconstruct it by hand.

The reason this matters is timing. The moment a firm most needs to see this picture is usually the moment it is hardest to assemble: a court's standing order asks what was verified, a client audit asks how AI was used on their matter, or a malpractice carrier asks for the firm's controls. Answering those questions today means digging through individual records after the fact. The dashboard is designed so the answer already exists as a view, continuously, rather than as a fire drill. The obligations it is built to map to are real and current, and worth naming plainly.

Why does this map to real obligations?

Court AI standing orders and the proposed amendment to Federal Rule of Civil Procedure 11 are converging on one expectation: a filer should be able to show that AI-assisted authorities were verified before they were filed. The proposed Rule 11 amendment is a proposal before the rulemaking process, not yet a rule, and should be treated as one — but the direction is consistent with the standing orders judges are already issuing. A representative example is the standing order from Judge Nina Wang of the District of Colorado, effective December 1, 2025, which requires counsel to certify that AI-assisted authorities are non-fictitious and have been reviewed by a human. That is an obligation to be able to demonstrate verification, not merely to have intended it.

ABA Formal Opinion 512 adds a distinct duty that a governance view is well suited to reflect. Opinion 512 addresses a lawyer's confidentiality and informed-consent obligations under Model Rule 1.6 when using generative AI — that is, the client-relationship and consent side of AI use. This is different from evidentiary privilege, and the dashboard should not conflate the two: recording that consent was captured under the Opinion 512 duty is not the same thing as an assertion about privilege in litigation. A dashboard that aggregates verifiable records against exactly these obligations is the difference between scrambling to reconstruct what happened and pointing to a trail that already exists.

How would it turn existing records into an evidence pack?

An evidence pack is the concrete output the dashboard is designed to produce on demand: a per-matter bundle of the records that already exist, gathered into a form a court, a client, or an insurer can check. The point of designing it as a pack rather than a screenshot is that every item in it stands on its own cryptography. Nothing in the pack asks the recipient to trust the firm's word or the vendor's word; each record can be verified independently against the transparency log it was sealed to. The steps below describe the intended flow from a live matter to a handable pack.

1. Read the records the platform already producedThe dashboard reads the signed certificates and attestations generated during the work — citation certifications, isolation attestations, consent captures. It does not regenerate or re-derive them; it reads what already exists.
2. Aggregate by matterRecords are grouped by matter so a firm sees, in one place, every AI touchpoint on a given case: which filings were certified, which interactions were isolation-attested, and what remains open.
3. Surface the open itemsAnything unresolved — a flagged citation, a held filing, consent not yet captured — is shown as open rather than hidden, so the trail reflects the real state of the matter, not an optimistic one.
4. Assemble the packOn demand, the relevant records for a matter are bundled into an evidence pack, each item carrying the inclusion proof needed to verify it against the log independently.
5. Hand it overThe firm gives the pack to whoever asked — a court, a client, a carrier — who can confirm the records are genuine and unaltered without trusting the firm or RankShield. The evidence stands on the cryptography, not on assurances.

What is the difference between a dashboard and a report you could already run?

A firm could, in principle, assemble a report today by collecting the individual certificates and attestations by hand. The reasons to build a dashboard instead are practical. First, timeliness: a hand-assembled report is stale the moment it is finished, while a live view reflects the current state of every matter continuously. Second, completeness: manual assembly is exactly where things get missed, and a missed open item in a governance report is worse than no report, because it reads as a clean bill of health that was never true. Third, verifiability: a conventional report is a document you are asked to trust, whereas the records the dashboard surfaces each carry their own proof.

That third point is the one that distinguishes this from ordinary business intelligence. Most dashboards summarize data you have to take on faith. This one is designed to summarize records that a third party can check for themselves — the summary is a convenience for humans, and the proof underneath is what a court actually relies on. If the dashboard and the underlying record ever disagreed, the record would win, because the record is the evidence and the dashboard is only a reader of it. That ordering is deliberate and is part of why the platform signs and seals at the moment of work rather than at the moment of reporting.

What would it store, and what would it never store?

The storage discipline for the dashboard is inherited from the platform, not invented for it. The records it reads store cryptographic digests, verdicts, and public metadata — the fact that a citation was certified, the result of that certification, the fact that an interaction was isolation-attested — and never the substance of a filing, a client's facts, or work product. That boundary is the reason an evidence pack can be public-checkable without becoming a disclosure risk: a court can confirm that the checks ran and have not been altered, without ever seeing what the filing actually said or how the case was argued.

The underlying records are signed with a composite post-quantum scheme — ML-DSA and SLH-DSA, the algorithms standardized as NIST FIPS 204 and FIPS 205 — and sealed to an RFC 6962 transparency log, the same append-only, tamper-evident structure used for certificate transparency across the web. Pairing a lattice-based and a hash-based signature is a conservative choice for evidence that may need to be checked years later. It is worth stating plainly that this is quantum-safe, not quantum-proof: it is designed to remain verifiable even if one algorithm family is later weakened, not to make any guarantee against every future attack. The dashboard adds a reading layer over these records; it does not change what they store or how they are signed. You can read more about that model on why verifiable.

Who is the governance dashboard built for?

It is built for the people inside a firm who are accountable for AI use but are not the person who ran any individual tool. A managing partner needs a firm-wide picture to answer for the practice as a whole. A general counsel or ethics partner needs to see where consent was captured and where it is still open, and to be able to demonstrate the firm's controls to a client or a regulator. A litigator responding to a specific standing order needs the evidence for one matter, fast, and in a form the court can verify. Each of those people needs a different slice of the same underlying trail, which is why the design centers on filtering a single source rather than producing separate reports.

The persona this is shaped around is the small and midsize firm — the practice that has adopted AI drafting to keep pace but does not have a dedicated litigation-support desk or a governance team to reconstruct an evidence trail on demand. Large firms can staff around this problem. The firms most likely to be caught flat-footed by a standing order or a client audit are the ones that cannot, and they are the ones a standing governance view is meant to serve. The goal is enterprise-grade accountability at a scale a small firm can actually operate: a view it switches on, not a department it has to build.

How does it connect to the rest of the platform?

The dashboard is a consumer of records produced elsewhere on the platform, and it is only as useful as those records are trustworthy. Citation certifications come from the certification workflow, described on citation certification, which resolves every cited authority before a filing is signed and issues a verifiable certificate. Isolation attestations come from privilege isolation, which records that an AI interaction ran inside an attested boundary. The approved-tool inventory the dashboard filters on is grounded in AI tool attestation, which establishes which tools a firm has sanctioned and attested. The dashboard's job is to gather those threads into one view — it does not replace any of them.

This is also why the honesty locks that govern each of those capabilities carry through to the dashboard. It cannot report a certification the certification layer did not produce; it cannot claim an interaction was isolated if no isolation attestation exists; and it does not originate legal conclusions about any of it. It is a mirror of the platform's records, and it is designed to be a faithful one — including reflecting what is missing. If you want to talk through how a governance view would fit a specific firm's workflow, the contact page is the place to start; the dashboard is being designed with input from firms that will use it.

Honest status: where does the governance dashboard actually stand?

This is a roadmap capability, and it should be read as one. The dashboard described on this page is not shipped. What is real and available today are the underlying records: the signed, sealed, independently verifiable certificates and attestations that the platform already produces as work is done. The reporting and evidence-pack layer — the cockpit that aggregates those records into a firm-wide view — is being designed, with input from design-partner firms, on top of records that already exist. We are describing a direction, not announcing a released product, and we would rather be plain about that than imply a screen exists that does not.

Two limits hold regardless of how far the dashboard develops. It reports evidence; it does not make a firm compliant. Aggregating verifiable records into a clean view does not discharge any obligation on its own — a lawyer still has to do the work, capture the consent, and make the judgments the rules require. And it does not render legal judgments: it will never tell a firm whether its AI use was permissible, whether a filing was adequate, or whether a duty was met. Those are questions for the lawyers and the court. The dashboard's role is narrow and deliberately so — to make the evidence of what was actually done easy to see and easy to verify. RankShield is a security vendor, and nothing here is legal advice.

What will the governance dashboard not do?

• MythIt makes your firm compliant with the AI rules.

  TruthIt does not. It reports verifiable evidence of what was checked and what is open. Compliance depends on the lawyer's work and judgment; the dashboard makes that work visible and provable, it does not perform it.

• MythIt decides whether your AI use was permissible.

  TruthIt renders no legal judgment. It does not opine on whether a filing was adequate, whether consent was sufficient, or whether a duty was met. Those determinations belong to the lawyer and the court, not to a reporting layer.

• MythIt stores your filings so you can review them there.

  TruthNo. The underlying records store digests, verdicts, and metadata — never the substance of a filing or privileged content. The dashboard reads those records; it does not become a second copy of your work product.

• MythIt guarantees an AI that never fabricates authority.

  TruthIt makes no such claim. It reports which citations were certified and which interactions were attested — the results, not a promise about the model that drafted anything. It cannot vouch for a generator it only reports on.

• MythIt is available now.

  TruthIt is on the roadmap. The signed, sealed records it would aggregate are produced today; the dashboard that presents them is being designed with design-partner firms and is not yet shipped.