What changed
FACT (headline only): New York courts adopted a new rule governing attorney use of AI — reported via a Ropers Majeski client alert surfaced in Google News. HYPOTHESIS: the rule imposes a disclosure/certification obligation on AI-assisted filings and effectively requires verification of AI-generated citations. The rule text was NOT provided in the input, so the exact obligation (a signed certification? a disclosure sentence? mere guidance?) is unverified inference.
Why now
Courts nationwide are reacting to sanctions cases over hallucinated citations (Mata v. Avianca and successors — background knowledge, not from provided sources). A NY-wide rule converts a reputational risk into a compelled compliance act for the largest state bar in the country. First-mover window exists only until the rule text is digested by incumbents; NY is also the template other state court systems copy.
Converging signals
(1) A state court rule on attorney AI use (FACT from headline); (2) a defined compelled class — attorneys filing in NY courts, ~190k registered (inference on size); (3) an existing e-filing rail (NYSCEF — inference) and an existing, well-documented pain: sanctions for fabricated citations. Rule + filer class + verifiable-artifact = the forced-filer convergence shape, though the middle leg (what exactly must be filed) is unconfirmed.
Customer pain
Attorneys who use AI in drafting now face two stacked risks: sanctions/embarrassment from hallucinated citations, and (if the rule says what the alert implies) a formal certification they must sign under their own name. Cite-checking is currently manual associate labor or expensive incumbent tooling; solo/small-firm lawyers — most of the NY bar — often have neither.
Who pays
Solo and small-firm NY litigators first (no procurement, card-swipe buyers), then mid-size firm practice groups per-seat. NOT the court system — this is a sell-to-the-filer play, explicitly not enterprise procurement.
Solved today
Westlaw Quick Check / LexisNexis brief analysis (expensive subscriptions), Clearbrief and similar legal-tech point tools, or an associate manually pulling every cite. Disclosure language, where required by individual judges' standing orders, is hand-drafted.
Why current solutions are bad
Incumbent tools are bundled into $200+/mo research subscriptions many solos don't carry; none are purpose-built around the specific NY rule's certification artifact; manual cite-checking is slow and error-prone precisely when AI drafted the text. A $10-per-brief 'verify + generate the certification' tool matches how a solo actually buys.
Proposed product
Web app + API: upload a brief (or paste text) → extract citations → verify each against CourtListener/Caselaw Access Project data (existence, caption match, quote-check) → flag unverifiable/hallucinated cites → emit a signed-ready AI-use disclosure/certification block worded to the NY rule. Per-filing pricing; per-seat for firms; API for legal-drafting tools that want compliance built in.
MVP version
Citation extractor (regex + LLM assist) + CourtListener API verification + a report card per citation + the NY-rule disclosure paragraph generator. No NYSCEF integration needed — output travels inside the attorney's own filing. 3-5 weeks of AI-assisted solo build; founder has capital for data/API costs.
30-day build
Week 1: obtain and read the ACTUAL rule/administrative order (this is the go/no-go gate). Weeks 2-4: build MVP against CourtListener; validate hallucination-detection accuracy on a corpus of known-sanctioned briefs; land 10 NY solo litigators as design partners via legal-tech communities.
60-day build
Public launch keyed to the rule's effective date; SEO/content play owning the query 'New York AI rule disclosure requirement'; CLE-adjacent webinar with a NY ethics attorney; per-filing Stripe checkout live.
90-day revenue plan
Target 150-300 paid filings/mo ($10-15 each) plus 20-40 seats at $49-99/mo ≈ $3-7k MRR. Then replicate the playbook to the next states/federal judges' standing orders — the 50-market expansion is the real prize.
Distribution path
Content/SEO on the rule itself (compliance-deadline queries convert), NY bar association channels and CLE circuits, legal-tech directories, and partnerships with solo/small-firm practice-management communities. Demonstrated-value motion: free scan that shows a hallucinated cite, paid certification output.
Pricing hypothesis
$10-15 per brief verified+certified (mirrors the founder's proven ELDT per-upload model), $49-99/seat/mo for repeat filers, usage-priced API for drafting-tool vendors.
Technical difficulty
Moderate. Citation parsing is a solved-ish problem (eyecite is open source); the hard part is coverage/recency of free case-law data vs. licensed Westlaw/Lexis data, and keeping false-negative rate (missed fake cites) near zero because the attorney signs based on your output.
Legal / regulatory risk
Not UPL (it's a verification tool, not legal advice), but there is real liability exposure if the tool blesses a fabricated citation and the attorney gets sanctioned. Needs tight disclaimers, E&O insurance, and conservative 'unverified' flagging. This is a product-design burden, not a licensing barrier — founder himself needs no license.
Platform dependency
Depends on case-law data sources (CourtListener/CAP free tier vs. licensed data). No platform owner can deplatform a tool whose output rides inside the attorney's own court filing.
Founder fit
Mixed. The SHAPE is his proven ELDT pattern (rule → compelled class → per-transaction fee) and he can build it fast. But it is NOT a portal-submission play — there is no third-party submission into NYSCEF, so his portal-integration moat doesn't apply — and legal tech is outside his industrial/public-records domain, in a market where credibility with attorneys matters.
Breakout potential
High if the rule is substantive: every state court system and dozens of federal judges are issuing AI standing orders — a rules-database + per-jurisdiction certification generator becomes a 50-state compliance product. Low if the rule turns out to be guidance or a one-sentence checkbox.
Final recommendation
CONDITIONAL PURSUE. This matches the founder's proven per-filing compliance shape and rides a genuine forced-buyer trigger, but the load-bearing fact — what the rule actually requires — is unverified inference from a headline. Spend 1-2 days retrieving the actual NY administrative order/rule text. GO if it imposes a concrete, recurring certification/verification duty on filings; PASS if it is guidance or a trivial disclosure, because the remaining product (generic cite-checking) is a crowded market where incumbents win.
Next action
Retrieve the actual rule text from the NY Unified Court System (nycourts.gov administrative orders / 22 NYCRR amendments), confirm: (a) exact obligation, (b) effective date, (c) which courts/filings covered, (d) required wording of any certification. That single document decides go/no-go.