Flycatcher Corp. Ltd et al. v. Affable Avenue LLC et al.

Flycatcher Corp. Ltd and Flycatcher Toys, Inc. v. Affable Avenue LLC, doing business as CJ DIST, et al., No. 24 Civ. 9429 (KPF) (S.D.N.Y.), before District Judge Katherine Polk Failla, a trademark/unfair-competition action (Flycatcher owns the "SMART SKETCHER" mark for children's educational toys; the Third Amended Complaint, filed May 23, 2025, alleged the defendants trafficked in stolen Flycatcher product). Two orders govern the AI-sanctions episode, and their consequences are distinct and must not be merged: (i) the Opinion and Order filed February 5, 2026 (Doc. 227, 33 pp.), cited by the court itself as Flycatcher Corp. v. Affable Ave. LLC, No. 24 Civ. 9429 (KPF), 2026 WL 306683 (S.D.N.Y. Feb. 5, 2026) (the "Sanctions Op."), which struck Affable's motion-to-dismiss papers and entered default judgment against Affable, and authorized a fee application, but fixed NO dollar amount; and (ii) the Opinion and Order filed October 9, 2026 (Doc. 364, 18 pp.), which quantified and awarded the fees and costs. Primary documents (both held in `files/sources/`, caption- and docket-verified on their faces): the Sanctions Opinion, Doc. 227: https://storage.courtlistener.com/recap/gov.uscourts.nysd.633195/gov.uscourts.nysd.633195.227.0.pdf (saved as `files/sources/D-176-flycatcher-sanctions-order-doc227.pdf`); the fee opinion, Doc. 364: https://storage.courtlistener.com/recap/gov.uscourts.nysd.633195/gov.uscourts.nysd.633195.364.0.pdf (saved as `files/sources/D-176-flycatcher-fee-order-doc364.pdf`). Note the fee opinion's caption differs from the sanctions opinion's: because default judgment had by then been entered against Affable and several co-defendants, the Clerk was directed to modify the caption (Doc. 364 lists Ozmos Company LLC first and terminates Affable and others from the docket). No Westlaw citation for Doc. 364 was located on the face of the document or verified from a primary source as of this draft; none is stated. Provenance: prepared October 21, 2026, in the case-monitoring lane as a backfill draft (the matter was flagged as a library gap in the October 18, 2026 sweep notes and confirmed still missing on October 21, 2026), after primary-source verification from the CourtListener/RECAP documents above and reconciliation against the Charlotin database (see (3) below). Timing note: the February 5, 2026 sanction was captured on first located reporting of its October 9, 2026 fee quantification (Law360, October 9, 2026; VitalLaw, October 15, 2026, both press reports verified against the orders and used for nothing else; the press description of the award as "nearly $39,000" in fees to the lawyers who uncovered the AI use matches the $38,877.59 awarded by Doc. 364).

Steven A. Feldman, counsel for Defendant Affable Avenue LLC, filed Affable's brief in support of its motion to dismiss the Third Amended Complaint on June 20, 2025 (Dkt. #153–156) "peppered with false citations." Counsel for co-defendant Top Experience Company LLC, Joel G. MacMull, investigated and reported the errors, first to Feldman (June 22, 2025), then to the Court (June 26, 2025, Dkt. #158): at least thirteen cited cases that did not exist, and eight cases that exist but do not contain the quotations attributed to them. The Court's own finding, adopting the figure Feldman himself supplied at the sanctions conference when he corrected the Court's "one-quarter" estimate, "Fourteen out of 60 cited cases", is that 14 of the 60 cases cited in the brief were fake: "The Court recognizes the mathematical truism that 14 out of 60 is less than one-quarter, but the fundamental point remains that it is 14 fake cases too many." The Court found the citations were AI hallucinations, "indeed, it is the only explanation the Court credits", produced by Feldman's research process, which the Court reconstructed from his sworn testimony at the August 22, 2025 sanctions conference (transcript at Dkt. #223): he located cases through Google, Google Scholar, and vLex (a research tool with AI components, offered through the New York State Bar Association), drawing also on a personal "repository" of cases originally compiled using Casetext with CoCounsel, access to which he lost at the end of December 2024 when Thomson Reuters acquired Casetext; he then ran the brief through three rounds of AI review, vLex, then a "cite checker," Paxton AI, then a further AI pass, and, by his own admission under oath, never cite-checked the brief himself ("The Court: You didn't fully cite check the brief, before you submitted it to me. Mr. Feldman: Correct, I did not."). The pattern then repeated twice more after he was on notice. First, his response to the Court's Order to Show Cause (Dkt. #164, filed July 11, 2025, one day late), a "conspicuously florid" submission quoting Ray Bradbury's Fahrenheit 451 and musing on the scribes of Ashurbanipal, contained a further false citation: a passage presented as a quotation from Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023), the foundational AI-hallucination sanctions decision, that *appears nowhere in Mata**; a Google search traced it to an October 24, 2023 article about Mata by attorney Christopher F. Lyon, which Feldman did not attribute. Feldman testified he used NotebookLM on the response (he said only to "review the content"); the Court found "AI generated this citation from the start" and disbelieved his explanations. Second, on August 8, 2025, while awaiting the sanctions hearing, after the Court's July 18, 2025 order (Dkt. #169) had denied him leave to withdraw and replace the brief and warned that he was "not excused from this professional obligation by dint of using emerging technology", Feldman filed a proposed reply brief (Dkt. #180) containing yet another nonexistent case: "Himmelstein [v.] Comcast of the D.C., LLC, 908 F.3d 49 (D.C. Cir. 2018), aff'g 44 F. Supp. 3d 1 (D.D.C. 2012)," neither of which exists; the case he meant was Himmelstein, McConnell, Gribben, Donoghue & Joseph, LLP v. Matthew Bender & Co., Inc., 37 N.Y.3d 169 (2021), a reported New York Court of Appeals decision he and others had already cited correctly elsewhere in the case. His August 12, 2025 letter (Dkt. #183) calling it a "clerical error" introduced "during a final verification" and pointing to a Google Scholar screenshot, which in fact showed a different case name and a different citation (Himmelstein v. Comcast of the Dist., L.L.C.*, 931 F. Supp. 2d 48 (D.D.C. 2013), an FCRA case that "makes no mention of" the New York statute under discussion), only "dug himself a deeper hole." Throughout, when the errors were first flagged, Feldman did not correct them: he told MacMull he was "unable to verify certain citations" because of his lost database access, expressed "regret" and said he would replace the citations but offered no timeline, asked MacMull to "share the list of results you compiled" (a list MacMull's email had already itemized) and to "run", i.e., cite-check, his corrected brief for him, and on June 24, 2025 emailed MacMull the false statement that he had "already notified opposing counsel and the Court that I am addressing the citation issues and will file a corrected memorandum as soon as possible"; the Court was in fact first notified by MacMull's June 26 letter.

The Court held that Feldman violated Federal Rule of Civil Procedure 11 "repeatedly and brazenly, despite multiple warnings from the Court and fellow counsel," and acted in subjective bad faith, the mens rea both Rule 11 (on the court's own initiative) and the court's inherent powers require, finding that "[h]e knew, or consciously avoided learning, that the cases he cited were not 'warranted by existing law,'" Fed. R. Civ. P. 11(b)(2); the problem "are not the use of AI per se, but rather Mr. Feldman's (i) knowing decision to use flawed methods of legal research and cite-checking; (ii) his inexplicable refusal to verify his submissions before filing them with the Court; and (iii) his unwillingness to come clean once these issues were revealed to the Court." (The opinion also notes New York Rule of Professional Conduct 3.3(a)(1); the sanctions rest on Rule 11 and the court's inherent powers, Chambers v. NASCO, Inc., 501 U.S. 32 (1991).) Weighing the five Second Circuit factors for terminal sanctions, intentional bad faith (yes), prejudice to the court and parties (yes: "significant resources investigating and responding"), pattern rather than isolated instance (yes: three submissions), correction (none: "He kept submitting erroneous citations and kept refusing to check his citations"; he "never 'corrected' his 'misconduct'"), and likelihood of further misconduct (yes: "no convincing plan to take ameliorative action"), and expressly considering and rejecting lesser sanctions, the Court imposed the following consequences, in two orders that must be kept distinct. February 5, 2026 (Doc. 227): (a) Affable's motion-to-dismiss submissions (docket entries 153 through 156) stricken; (b) default judgment entered against Affable Avenue LLC in favor of Plaintiffs, a case-terminating sanction imposed on the client for its counsel's conduct (the court citing Link v. Wabash R.R. Co., 370 U.S. 626, 634 (1962): a party "cannot 'avoid the consequences of [its lawyer's] acts or omissions'"); damages assessment deferred until disposition as to the remaining defendants (no damages figure for Affable appears in either order verified here); (c) MacMull granted leave to apply for attorneys' fees against Feldman personally under Rule 11 and/or 28 U.S.C. § 1927, the Court finding that Feldman had "multiplie[d] the proceedings in [this] case unreasonably and vexatiously," § 1927, and ordering the application filed within 30 days. October 9, 2026 (Doc. 364): on MacMull's fee petition (filed on behalf of his then-client Top; awarded to MacMull and his firm, Mandelbaum Barrett PC, after his withdrawal as Top's counsel), the Court awarded $38,877.59 in attorneys' fees and costs, payable by Feldman personally within 30 days of the opinion unless the two agreed a different schedule, composed of $37,881.00 in fees (MacMull and his partner Brian Block billed 73.2 hours at a discounted $575/hour, 24.1 hours responding to the misconduct, 37.1 hours on the opening fee brief, 12 hours on the reply, for $42,090.00 sought, reduced by a 10% across-the-board reduction the Court imposed "largely motivated by Mr. Feldman's limited ability to pay": "the Court does not accept this as an excuse for his misconduct… it recognizes that for Mr. Feldman, a solo practitioner who has indicated a certain degree of financial limitations, a roughly $40,000 penalty is substantial") plus $996.59 in Westlaw research costs awarded in full; the total sought had been $43,086.59. Neither order imposes any further condition on Feldman that was located in these documents (no CLE, certification, filing restriction, or disciplinary referral appears in Doc. 227 or Doc. 364). Charlotin reconciliation: Charlotin's database carries this matter in two rows. Row "(2)," Flycatcher v. Affable Avenue (2), dated 2026-02-05, court "S.D. New York," party "Lawyer," AI tools "NotebookLM; vLex; Paxton AI," outcome "Brief Struck; Default Judgment; Adverse Costs Order," monetary penalty "38878 USD," professional sanction "No"; its four hallucination items (14 of 60 citations fake; the nonexistent Himmelstein citation; the false Mata quotation traced to the Lyon article; the Google Scholar misattribution) all match the Sanctions Opinion, and his held source document for that row is the same Doc. 227. Row "(1)," dated 2025-07-18, records the Court's interlocutory July 18, 2025 order (Dkt. #169) with outcome "Decision on sanctions reserved" and AI tool "Unidentified", consistent with Doc. 227's account of Dkt. #169 (the Dkt. #169 order itself was not pulled for this draft; its contents are verified only as described in Doc. 227). Discrepancy: Charlotin attaches the "38878 USD" monetary penalty to the February 5, 2026 row, but the February opinion fixed no amount, it only authorized the fee application and set a 30-day deadline; the $38,877.59 figure (his 38878 is that award, rounded) was not determined until the October 9, 2026 opinion (Doc. 364), which Charlotin does not carry as a separate row. His outcome fields (brief struck; default judgment; adverse costs order; no professional sanction) otherwise match the primary record. Primary documents govern.

R5, Denied or Doubled Down (graded from the two court orders only). The February 5, 2026 opinion records: a false statement to opposing counsel that he had "already notified opposing counsel and the Court" about the citation issues when he had not; promised corrections that were never made and offered on no timeline; a show-cause response, filed a day late, containing a new fabricated quotation attributed to Mata; and a further filing with another nonexistent case submitted while the sanctions hearing was pending and after express warnings, after which the Court found he "remained unable to accept full responsibility and admit his mistakes," that his answers under oath were "often entirely unresponsive," and that although he "at times purported to claim full responsibility for the errors (see Tr. 52 ('I'm saying they are my fault regardless.')), he undermined this acceptance of responsibility by repeatedly minimizing his behavior and offering improbable explanations", minimizing included correcting the Court's arithmetic ("Fourteen out of 60 cited cases"). The October 9, 2026 opinion extends the record past the sanction: invited to resolve the fee award informally, Feldman "gave Mr. MacMull the runaround," demanded an invoice already sent and a "single, proper invoice in standard form" while offering $1,500 monthly installments, and prompted the Court's conclusion that "Mr. Feldman has not, and apparently cannot, learn from his mistakes." The fee opinion likewise characterizes the February record as submitting AI-produced false citations "and then lying about their genesis" (Sanctions Op., at *12, as quoted in Doc. 364).

---

Every stage of this cascade was preventable with the most basic verification step: read the case before you cite it. The Second Circuit rule the Court applied, from Park v. Kim, 91 F.4th 610, 615 (2d Cir. 2024), is that Rule 11 at minimum requires attorneys to "read, and thereby confirm the existence and validity of, the legal authorities on which they rely", "[a] fake opinion is not 'existing law.'" Three specifics deserve emphasis because each is a documented failure mode in this record. First, an AI "cite check" is not a cite check: Feldman's process was to have AI review AI-assisted research (vLex, then Paxton AI, then another AI pass), and the Court found that process itself introduced or ratified the errors; verification means pulling the authority itself from a reporter, PACER, CourtListener, or another primary source and reading it, a step available to Feldman without any paid database, as his own later PACER downloads for the reply brief show, and as the Court put it, "If Mr. Feldman could not verify a citation, he should not have cited it." Lost Westlaw access is an explanation for inconvenience, not a defense: free sources would have caught a citation, "908 F.3d 49," that resolves to an unrelated Fourth Circuit criminal appeal, and a D.D.C. cite offered for a New York statute the whole case had already cited correctly at 37 N.Y.3d 169. Second, the response is the sanction multiplier: opposing counsel handed Feldman an itemized list of errors and a day to self-report; a prompt withdrawal, a corrected brief on a stated timeline, and a candid letter to the court in June 2025 would have presented the Court a correctable mistake. Instead came a false claim of notification, a request that opposing counsel do the cite-check, a late show-cause response with a fabricated Mata quotation (of all cases), and a third bad filing while a sanctions hearing was pending, the sequence that converted Rule 11 error into a bad-faith finding and a terminal sanction. Third, the client pays for the lawyer's refusal: default judgment fell on Affable, which lost its motion and its defense on liability without any merits adjudication, because a client is bound by its lawyer-agent's acts. A firm or solo practitioner with no verification infrastructure has a client-protection problem, not just a personal one.

• storage.courtlistener.com: https://storage.courtlistener.com/recap/gov.uscourts.nysd.633195/gov.uscourts.nysd.633195.227.0.pdf
• storage.courtlistener.com: https://storage.courtlistener.com/recap/gov.uscourts.nysd.633195/gov.uscourts.nysd.633195.364.0.pdf