Accountability-as-a-Service
Content submitted to our platform, verified against primary publications from regulators, courts, and governmental bodies.
The Hallucination Register is our review service on the content submitted to us. Content is submitted to our platform by (a) users submitting their own work products for review, (b) users submitting content they have received from others, (c) users submitting specific content pieces they suspect to be a case of hallucination, for us to check, (d) public-facing web sources accessible via ungated web searches, and (e) AI systems sending us their outputs.
Each submitted item is checked against the actual publication from the primary regulatory, legal, or governmental body. Where we find that the submitted content materially differs from what the primary body has actually published, we flag it and record it in this Register with a unique citation ID.
The purpose of the Register is to give users the capability to flush out such identified instances from their critical outputs, before those outputs are used in professional or public-facing settings.
The Register does not catalogue errors in the publications of regulators, courts, or governmental bodies. Instead, the Register relies on the authenticity of those publications to flag material discrepancies in the content that is submitted to us.
We provide links to make it easy for the readers to refer to the original publications of regulators, courts, or governmental bodies as the authentic source of information.
The Hallucination Register operates at three commercial surfaces. Each carries a different function. All three run against the same verification standard: content is checked against the actual publication from the primary regulatory, legal, or governmental body.
Surface 1
Our production discipline
Before any RegLegBrief output publishes, it is checked against the Register. Our briefs, our work products, our responses to user queries — all pass through the same verification standard that governs the Register itself. This is how we stand behind the content we deliver, and it is part of what users pay for when they subscribe to our services.
Surface 2
The public catalogue
The Register is published on this platform as a catalogue of confirmed hallucinations. Each entry carries a citation ID, the jurisdiction, the primary body whose publication was misquoted, and a broad domain tag. Users browsing the catalogue identify entries relevant to their areas of practice; subscribers at Reader Pro and above can access the full analytical detail on each entry. The catalogue is updated as new entries pass verification.
Surface 3
The review service
Users submit work products — their own, their counterparties', their vendors', or outputs from automated tools their business relies on — and we check them against the Register, as well as our internal database, and further against the primary publications from the referenced bodies. Confirmed hallucinations in the submitted material are flagged with the corresponding citation IDs, so the content can be corrected before it reaches court, a client, a regulator, or a board. The review service is priced per submission, at the tier appropriate to the material.
Four entries as of 21 April 2026. Each entry carries a citation ID, a verification date, the jurisdiction, the primary body whose publication was misquoted, a broad domain tag, and a status. The specific hallucinated content, the verification source, and the user advisory on each entry are available to Reader Pro subscribers and above. Filter and search controls will arrive as the catalogue grows.
The full detail of this entry — including the specific hallucinated content, the verification source, and the user advisory — is available to Reader Pro subscribers and above.
Upgrade to Reader Pro to unlock the full Register, including the detail on this entry and on every other confirmed hallucination in the catalogue.
Upgrade to Reader Pro →The full detail of this entry — including the specific hallucinated content, the verification source, and the user advisory — is available to Reader Pro subscribers and above.
Upgrade to Reader Pro →The full detail of this entry — including the specific hallucinated content, the verification source, and the user advisory — is available to Reader Pro subscribers and above.
Upgrade to Reader Pro →The full detail of this entry — including the specific hallucinated content, the verification source, and the user advisory — is available to Reader Pro subscribers and above.
Upgrade to Reader Pro →Full analytical detail on each entry is available to Reader Pro subscribers and above. This includes the specific hallucinated content, the verification source, and the advisory users need to check their own work against the entry.
Think something is a hallucinated claim in circulation? Submit it for review. Confirmed submissions join the Register with a citation ID, and the submitter's profile receives a Confirmed Case badge carrying that ID. Every confirmed submission helps other users flush similar hallucinations from their own critical work.
In scope
Content that misquotes or fabricates publications from regulators, courts, or governmental bodies, where the content is materially discrepant from what the primary body has actually published. Examples: a cited case that does not exist in the court's own records, a regulatory threshold figure that does not match the regulator's own publication, a claimed rule or notice that the regulator has not issued, a quotation attributed to a court or regulator that the primary source does not contain.
Not in scope
Errors within a regulator's, court's, or governmental body's own publication. The Register does not catalogue the primary body's own material; it relies on the primary body's publication as the verification standard. Opinion content, commentary, and general factual errors unrelated to regulatory, legal, or governmental publications are also outside scope.
Check your email to verify the address and complete the submission. Until verification, the submission will not enter the review queue.
The verification link may have expired.
Your submission is in the review queue. If your submission is confirmed as a hallucination, the citation ID will be added to your profile as a Confirmed Case badge. You can view your profile at any time via the platform dashboard.
On clicking Submit, a verification email is sent to the address you provided. Your submission is held until you verify the email. Verification is a single click inside the email; it marks your submission as ready for the review queue.
Once verified, your submission enters the review queue. Submissions are reviewed against the actual publication of the primary body. If the submission is confirmed as a hallucination, it is added to the Register with a new citation ID, you are notified by email, and the citation ID is added to your profile as a Confirmed Case badge. If the submission does not meet the confirmation standard, you are similarly notified.
Review times vary. The Register is maintained through automated verification against primary publications, and some submissions take longer than others to resolve.
Verifying your email address also registers you as a free-tier user on the platform, with access to the Hallucination Register teaser catalogue and the platform's other free-tier features.
The paid review service on your own work.
Hallucination Check is the review service that applies the platform's verification discipline to your own content. You submit a document; we check it for hallucinations; we return the findings to you. Your document is removed from platform storage once the review is delivered and your access window closes.
The service runs two layers of verification on every submission.
Against the Hallucination Register. Your document is checked against every confirmed entry in the platform's Hallucination Register. Matches are flagged with the corresponding citation IDs, so the specific hallucinated items that have contaminated your content are named directly.
Against the relevant primary sources. Your document is then verified end-to-end against the actual publications of the primary regulatory, legal, or governmental bodies referenced in it. Material discrepancies identified through Layer 2 are flagged with the primary source reference and the correct position. Layer 2's verification extends beyond what the public Register alone contains. Hallucinations confirmed through our ongoing work, but not individually citable as public Register entries, remain part of our internal research and strengthen every review we conduct.
You submit a document — your own draft filing, your proposal to a committee, your memorandum, your advisory note, your marketing content, your disclosure, your AI-drafted first pass. Content in any form you will rely on professionally, or that will reach a client, a court, a regulator, or a board.
What you receive by default is two things delivered together: the marked-up document, your original returned with flagged passages highlighted, commented, and linked to the verification finding on each flag; and the audit report, a structured document listing every finding with its citation ID, the primary source against which the discrepancy was identified, and the recommended remediation. Suitable for filing alongside your work as a record of verification.
If you only need the audit report without the marked-up document, the submission carries a 50% discount. Marked-up output is delivered for DOCX and plain-text submissions at launch. PDF submissions are returned with the audit report only at the 50% price point; marked-up output for PDF is on the build roadmap.
Priced by the document. The standard unit is one check up to 1,000 words at $5. Longer documents scale linearly. Each subscription tier includes a monthly free allowance.
| Tier | Monthly free allowance | Equivalent word budget |
|---|---|---|
| Reader Pro | 1 check | 1,000 words |
| Professional | 3 checks | 3,000 words |
| Executive | 5 checks | 5,000 words |
| Corporate API | Custom per engagement | — |
Before any review initiates, you see your submission's word count, standard-check equivalent, remaining free allowance, and overage cost. You confirm before the review runs. Submissions beyond the free allowance are billed at the standard rate of $5 per 1,000-word unit.
Every review is delivered within two business days of submission acceptance. Hallucination Check is a thorough verification, not a fast service. Every submission is run end-to-end against the relevant primary publications from regulators, courts, and governmental bodies. The verification discipline sets the pace, not the clock.
Your submitted document is reviewed by the platform to produce your Hallucination Check output. Once the output is delivered, your document remains accessible to you for seven calendar days through your account, and is then permanently removed from platform storage.
The platform does not retain your document beyond that access window, does not store it for future reference, and does not hold it for any purpose beyond the single review that was commissioned.
What the platform retains is its own work: the hallucinations identified and verified through the review, and the internal audit trail of where and how each hallucination was found. Hallucinations that exist in the public information environment are catalogued in the Hallucination Register. Hallucinations that surface within your private content and are not otherwise in public circulation remain in your review alone, and are not catalogued in any Register attributable to you.
Individual practitioners, boutique firms, and corporate teams. Professionals producing content that carries their reputation. In-house counsel, compliance officers, heads of tax, data protection officers, and company secretariats reviewing work before it reaches the board or the regulator. Advisory teams reviewing drafts before delivery to clients. Professionals whose content will be relied on downstream, where the cost of a hallucination slipping through is greater than the cost of the check.
Corporate-scale engagements are configured individually per client, with pricing, volume, integration, and turnaround shaped around the specific operational workflow.
The Hallucination Register is one of several professional efforts addressing the quality of regulatory and legal content circulating today. Each of the following resources tracks events, research, or responses in the space. Each is architecturally different from the Hallucination Register, which catalogues specific hallucinated content items for user-checkability. Links below were verified on the date shown; we do not maintain ongoing checks on external resources.
AI Hallucination Cases Database
An investigative tracker of court cases in which hallucinated content has been identified. Over 1,330 cases catalogued across jurisdictions as of early 2026. The database focuses on documented events — court orders, sanctions, and judicial findings — rather than on the specific fabricated claims themselves.
Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools
Published in the Journal of Empirical Legal Studies (2025). The first preregistered empirical evaluation of commercial legal-AI research tools. Found that leading products from LexisNexis and Thomson Reuters hallucinated between 17% and 33% of the time. Establishes that retrieval-augmented generation does not eliminate hallucination risk.
Law360 Pulse AI Tracker
An ongoing tracker of federal judicial standing orders on the use of AI in court filings. Covers approximately 2% of US federal district and magistrate judges who have issued standing orders, and the cases driving those orders. Useful for understanding how the judiciary is responding to AI-generated content in filings.
RAILS AI Use in Courts Tracker
An open-source tracker of court orders, local rules, and judicial guidelines on AI use, covering the US and other jurisdictions. Downloadable data, searchable filters, and classification by document type. Maintained as an academic-judicial resource rather than a commercial one.
Ropes & Gray Court Order Tracker
An interactive tracker of standing orders, local rules, and judicial decisions on AI use in legal practice. Published by a leading international law firm and maintained as a professional resource for practitioners navigating cross-jurisdictional AI-use requirements.
Each of the resources above serves a distinct purpose: documenting cases, evaluating tools, tracking judicial responses. The Hallucination Register takes a different architectural approach: it catalogues the specific hallucinated content items themselves — the fabricated case citations, the misquoted regulatory thresholds, the fictitious attributions — so that professionals can check their own work products against confirmed hallucinations before those products are relied on. The two-layer Hallucination Check service extends this capability to direct review of client content.
The combination of efforts, across researchers, publishers, law firms, and platforms, represents a shared professional response to a shared professional problem.
The Hallucination Register reflects our verification work against the publications of primary regulatory, legal, and governmental bodies. It is a catalogue of confirmed material discrepancies between content submitted to our platform and the actual publications of those bodies.
The Register does not catalogue errors in the publications of regulators, courts, or governmental bodies. Those publications are the verification standard we check against. Where a Register entry identifies a discrepancy, the original primary publication remains the authoritative reference, and we direct readers to it.
We catalogue what has been verified. We do not advise on what users should do about it. Primary publications remain the authoritative reference; professional judgement remains the user's.