The CFTC Prediction Markets Comment Record

PMP retrieved comments filed in response to the ANPRM (RIN 3038-AF65) from the CFTC's public-comment portal at comments.cftc.gov. Substantive filings (314) were analyzed individually or assigned to coordinated coalition efforts. PDF and DOCX exhibits were retrieved for every attachment-bearing filing. Five of the substantive filings were first indexed by the CFTC under a different Federal Register notice and later re-posted to this docket under new comment IDs; this page and the download cite the current IDs, and the download notes each former ID. Form-letter and template-driven filings — including the More Perfect Union anti-PM template and the Kalshi AI comment portal output (~1,865 filings, ~53% of the full docket) — were tagged at the cluster level rather than classified individually. Data current as of May 27, 2026.

Each of the 314 substantive comments — defined as comments with original analytical content, signed by an identifiable individual or organization, typically with a PDF or DOCX attachment — was processed through PMP's AI-assisted classification pipeline, with human review on the highest-weight filings and any low-confidence calls. Template-conformant coalition members (Acoma/Stillaguamish KalshiEX-litigation, Lincoln-Feinstein 2010 Colloquy, IGA templates) were catalogued via the template anchor — read in full — and members tagged at the cluster level. The ~2,500 mass-campaign comments — including ~1,865 generated by the Kalshi AI comment portal and ~555 More Perfect Union template filings — were similarly tagged at the cluster level after the template signature was identified; not every comment in those clusters was read in full. The ~710 short individual unaffiliated comments were skimmed in aggregate to confirm the absence of new analytical material; they were not classified into one of the five buckets individually.

Each substantive comment was assigned a primary bucket (one of the five) and, where applicable, a secondary bucket.

Mass campaigns identified via shared opening-text signature on the first ~200 characters of the inline comment text. The operator-portal AI cluster was identified via the template signature "Dear Chairman and Commissioners, My name is [X], and I'm a [role] based in [State]." A user-submitted screenshot of an exchange operator's customer portal interface confirmed the AI-tool drafting flow on April 29; a Dan Bernstein X thread on April 28 had independently documented the same "Make Your Voice Heard / Build Your Comment" landing page. To PMP's knowledge, this is the first documented instance of an exchange operator using AI-generated retail-customer mobilization as a primary docket-engagement strategy — and on a May 27 sweep of the full 3,534-comment record, 1,865 of 1,930 mass-block filings (96.6%) matched the Kalshi-portal template, accounting for ~53% of the entire public docket. Another 47 filings in that block carry the portal's salutation but were shortened or rewritten by the filer, in several cases to oppose the operator; they are tagged separately and are not counted in the 1,865. Two in-docket filings explicitly disclose Kalshi-portal authorship (Bader 117562, Weinberger 115831). Five filings in the record are signed "Test User" and carry identical portal-template text (115285, 115677, 115678, 115679, 115684). Both clusters are tagged at the cluster level rather than counted individually toward the substantive bucket distribution.

PMP does not score the technical merit of legal arguments in any comment. PMP does not predict NPRM language. PMP does not recommend specific regulatory outcomes. The goal of this analysis is structural — who commented, what they argued, and where the coalitions are. Editorial judgments are flagged with the prefix "PMP's read."

PMP designed the bucket classifications and editorial framing. Parts of PMP's analytical pipeline — template-cluster detection and classification scaffolding — were AI-assisted. The authoritative record of any comment is the document filed at comments.cftc.gov; PMP's classifications and excerpts are interpretive and should not be quoted in place of the source. When quoting a specific comment, link to the original at comments.cftc.gov. To cite PMP's analysis itself (the matrix, classifications, structural read), use the citation in the footer below.

Filers, journalists, and other parties who identify classification errors can reach us via the contact page.