Braica v. Frankowski, No. 3:24-CV-1709 (D. Conn. 2025)

Case details
Full caption
Braica v. Frankowski (Anthony Braica v. Tom Frankowski)
Country
United States
Jurisdiction
Federal
Court
District of Connecticut (D. CONN)
Decided
2025
Disposition
Dismissed
UNITED STATES DISTRICT COURT DISTRICT OF CONNECTICUT --------------------------------------------------------------- x : : : : : : : : : x MEMORANDUM & ORDER 3:24-CV-1709 (VDO) ANTHONY BRAICA, Plaintiff, -against- TOM FRANKOWSKI, Defendant(s). --------------------------------------------------------------- VERNON D. OLIVER, United States District Judge: Plaintiff Anthony Braica repeatedly submitted briefs to this Court that were strewn with generative artificial intelligence (“AI”) hallucinations. Upon uncovering fabricated case law, fictitious quotes, and repeated misstatements of law in Plaintiff’s briefs, the Court issued an order to show cause why sanctions should not be imposed under Rule 11(b) of the Federal Rule of Civil Procedure (the “Order to Show Cause”). For the reasons that follow, the Court ADMONISHES Plaintiff and STRIKES the briefs containing AI-generated hallucinations. I. BACKGROUND Plaintiff Anthony Braica, who was serving a period of probation connected to a state court sentence, brought this action against probation officer Defendant Tom Frankowski, alleging Defendant sexually assaulted him while conducted a search to confirm he was not using a device to falsify a drug test.1 Plaintiff alleges five claims in his operative amended complaint (the “Amended Complaint”): (1) violation of his Fourth Amendment right to bodily privacy; (2) cruel and unusual punishment in violation of the Eighth Amendment; (3) retaliation in violation of the First Amendment; (4) intentional infliction of emotional distress; 1 Am. Compl., ECF No. 25. Case 3:24-cv-01709-VDO Document 35 Filed 12/15/25 Page 1 of 12
2 and (5) negligent infliction of emotional distress.2 On August 18, 2025, Defendant filed a partial motion to dismiss Plaintiff’s second and fifth claims: cruel and unusual punishment in violation of the Eighth Amendment, and negligent infliction of emotional distress (the “Motion to Dismiss”).3 On September 8, 2025, Plaintiff filed a response to the partial motion to dismiss (the “Response”).4 The Response was inundated with AI-generated hallucinations. Most often, these hallucinations came in for the form of misstatements of law, meaning Plaintiff distorted or materially altered the holdings of real cases.5 In one instance, however, he included a fictitious quote from an actual case.6 And in another, Plaintiff went as far as to include an entirely fabricated case.7 After conducting a review of the Response, the Court found the following AI-generated hallucinations: State v. Kelleher (citation nonexistent). This is an entirely fabricated case. The citations Plaintiff lists lead to different cases unrelated to the instant matter.8 2 Id. 3 Mot. to Dismiss, ECF No. 29. 4 Resp., ECF No. 30. 5 See id. 6 ECF No. 30. at 5. 7 Id. at 4. 8 Id. Plaintiff’s full citation reads “State v. Kelleher, 181 Conn. 618, 627, 879 A.2d 222 (2005) held ‘the attachment of a GPS device to a person, and the ongoing collection of location data, constitutes a search with the Fourth Amendment.” Id. The citations Plaintiff lists—181 Conn. 618 and 879 A.2d 222—lead to real cases that are not called State v. Kelleher and have no relevance to the instant matter. Case 3:24-cv-01709-VDO Document 35 Filed 12/15/25 Page 2 of 12
3 Martin v. Brady, 802 A.2d. 814 (Conn. 2002). Plaintiff attributes a wholly hallucinated quote to this case.9 Solem v. Helm, 463 U.S. 277 (1983). Plaintiff cites this case to rebut Defendant’s argument that the Fourth Amendment—not the Eighth Amendment—governs Plaintiff’s claim because Plaintiff was on probation at the time of the incident. Solem is entirely irrelevant to this argument. In Solem, the Eighth Amendment violation occurred while the individual was incarcerated, not on probation, and the case has no other relevance to the instant matter. Graham v. Florida, 560 U.S. 58 (2010). Like Solem, the violation in Graham occurs while the individual is incarcerated, not serving a term of probation.10 State v. Moore, 963 A.2d. 1019 (Conn. App. 2009). Plaintiff also cites Moore to rebut Defendant’s argument that the Fourth Amendment governs Plaintiff’s claim. Yet this case examines whether a search of an individual’s apartment by his probation officer violated his rights under the Fourth Amendment, not the Eighth Amendment. Further, a Plaintiff attributes an entirely fabricated quote to this case.11 Rochin v. California, 342 U.S. 165 (1952). Plaintiff also cites this case to refute Defendant’s Eighth Amendment argument, but this case involves the Thirteenth and Fourteenth Amendments, not the Eighth.12 Steffel v. Thompson, 415 U.S. 452 (1974). Plaintiff uses this case to again discuss the Eighth Amendment, but it concerns an individual’s right under the First and Fourteenth Amendments, not the Eighth.13 United States v. Jones, 565 U.S. 400 (2012). Plaintiff cites this case for the proposition that “the attachment of a GPS device to a person” is a Fourth Amendment search. 14 As explained below, the use of a GPS device has no relevance to the instant case. Even leaving that aside, Plaintiff directly misstated the law because this case concerns the attachment of a device to a vehicle, not a person. Quon v. Arch Wireless Operating Co., Inc., 529 F.3d 892 (9th Cir. 2008). Plaintiff cites this case in relation to his misdirected GPS-monitoring argument, but he fails 9 ECF No. 30 at 5. 10 Id. 11 Id. at 3. 12 Id. 13 Id. 14 Id. Case 3:24-cv-01709-VDO Document 35 Filed 12/15/25 Page 3 of 12
4 to mention it was explicitly reversed and remanded by the Supreme Court of the United States.15 Hamilton v. Lajoie, 660 F. Supp. 2d 261, 265 (D. Conn. 2009). Plaintiff cites this case for the proposition that sovereign immunity does not shield government employees from liability for wanton, reckless, or malicious acts. While this is generally true, Plaintiff fails to acknowledge that the Hamilton court dismissed the negligence claims as barred by the doctrine of sovereign immunity.16 In addition to AI-generated misstatements of law and fabricated quotations, Plaintiff’s response contains sections entirely unrelated to the Amended Complaint or Motion to Dismiss. This deviation into an unrelated area of law is also highly indicative of unvetted AI use. Plaintiff dedicates an entire section of his Response to discussing whether GPS monitoring implicates the Fourth Amendment—an issue that is not before the Court in any form at this point in the case.17 Up until this brief, Plaintiff never mentioned GPS monitoring, but nevertheless this discussion comprises the bulk of the Response.18 Separately, he also broaches the issue of injunctive relief despite not requesting such relief in his Amended Complaint, or at any other point in the case.19 Defendant pointed out these deficiencies in his reply brief, but did not formally move for sanctions (the “Reply”).20 Then, on September 15, 2025, Plaintiff attempted to explain his errors by filing an improper sur-reply without seeking leave of the Court (the “Sur-Reply”).21 15 Id. at 4. 16 Id. 17 Id. at 3–5. 18 Id. 19 Id. at 2, 5–7. 20 See generally Reply, ECF No. 31. 21 Sur-Reply, ECF No. 32. See L.R. Civ. Pro. 7(d) (“No sur-replies may be filed without permission of the Court, which may, in its discretion, grant permission upon a showing of good cause.”). Case 3:24-cv-01709-VDO Document 35 Filed 12/15/25 Page 4 of 12
5 In the Sur-Reply, Plaintiff both stood firm on his false assertions and conceded to making some errors. For example, he maintained that although Solem and Graham do not stand for the proposition he claimed, that they nonetheless “remain good law[,] have not been overturned, [and] stand for the proposition that punishment must be proportionate to the offense.”22 He admitted to a possible “misattribution” of the quotation he credited to the court in Moore, but simultaneously continued to submit AI-generated hallucinations to the Court.23 For instance, Plaintiff again cited an entirely fabricated case and continued to cite quotations and cases that have no relationship to his argument that the Eighth Amendment applies with equal force to incarcerated and non-incarcerated individuals.24 Once again, the Court recounts the issues below: Sheppard v. Roberts (citation nonexistent).25 This is an entirely fabricated case. The citations Plaintiff lists lead to different cases unrelated to the instant matter. Rochin v. California, 342 U.S. 165 (1952). Plaintiff again cites Rochin, this time to reference its “shocks the conscience” standard,26 but as discussed above, this case analyzes the Thirteenth and Fourteenth Amendments. It is not relevant to Plaintiff’s Eighth Amendment claims. Fuller v. Comm’r of Correction, 815 A.2d 208 (Conn. App. 2003). Plaintiff cites this case for the proposition that individuals on probation are protected by the Eight Amendment.27 While this case concerns the Eighth Amendment, it has little, if anything, to do with the rights of individuals on probation. Rather, it dismisses a 22 ECF No. 32 at 2. 23 Id. 24 See generally id. 25 Id. at 4. Plaintiff’s full citation reads “Sheppard v. Roberts, No. 3:19-CV-1345 (JCH), 2021 U.S. Dist. LEXIS 133288, at 12 (D. Conn. July 16, 2021).” Id. This is a fabricated case. The cases located at the citations he lists—3:19-CV-1345 and 2021 U.S. Dist. LEXIS 133288—have entirely different case names and are unrelated to the instant matter. 26 Id. at 2 27 Id. Case 3:24-cv-01709-VDO Document 35 Filed 12/15/25 Page 5 of 12
6 claim that the Eight Amendment requires prison housing classification systems that separate violent and nonviolent inmates. On October 3, 2025, after identifying these repeated hallucinations in Plaintiff’s briefs, the Court issued an Order to Show Cause why sanctions should not be imposed.28 In Plaintiff’s Response to the Order to Show Cause, he admitted to using publicly available websites to assist with legal research and drafting, but stopped short of specifically admitting to using generative AI.29 He apologized the Court and stated that he “wishes to be fully candid,” yet he failed to precisely identify the source of the hallucinations he cited. Further, Plaintiff admitted to not reading or reviewing the briefs: Lacking legal training, Plaintiff did not possess the skills necessary to independently verify the accuracy of the information generated by the websites. I accepted the output as correct an incorporated it into [my] briefing in good faith, not knowing it was riddled with falsehoods and misrepresentations which I [should’ve].30 The Response to the Order to Show Cause concludes by listing corrective actions Plaintiff “solemnly commits” to following, such as (1) “ceasing use of misinformed websites,” (2) “diligently verifying every legal citation and quotation,” (3) adhering to the “Federal Rules of Civil Procedure and the Local Rules of this Court,” and (4) tailoring all future filings “to the specific legal and factual issues relevant to the claims at hand.”31 However, it stops short of explaining how he plans to fulfill this list of obligations. 28 See O.S.C., ECF No. 33. 29 Resp. to O.S.C., ECF No. 34 at 2. 30 Id. at 2–3. 31 Id. at 4. Case 3:24-cv-01709-VDO Document 35 Filed 12/15/25 Page 6 of 12
7 II. LEGAL STANDARD Litigants, including those proceeding pro se, are obligated to ensure their arguments are warranted by existing law or otherwise legally tenable. Under Rule 11 of the Federal Rules of Civil Procedure, a litigant presenting a filing to a court “certifies that to the best of the person’s knowledge, information, and belief, formed after an inquiry reasonable under the circumstances . . . the claims, defenses and legal contentions are warranted by existing law or by a nonfrivolous argument for extending, modifying, or reversing existing law or for establishing new law.” Fed. R. Civ. P. 11(b). By its express terms, Rule 11 applies to represented and pro se litigants alike. Maduakolam v. Columbia Univ., 866 F.2d 53, 56 (2d Cir. 1989). “Although courts may afford them greater leniency, pro se litigants are not immune from Rule 11 sanctions.” Ferguson v. Comm’r of Tax & Fin., 739 F. App’x 19, 22 (2d Cir. 2018). A court may sanction a litigant under Rule 11 for “among other things, misrepresenting facts or making frivolous legal arguments.” Muhammad v. Walmart Stores E., L.P., 732 F.3d 104, 108 (2d Cir. 2013) (per curiam). “At the very least, the duties imposed by Rule 11 require that [litigants] read, and thereby confirm the existence and validity of, the legal authorities on which they rely.” Park v. Kim, 91 F.4th 610, 615 (2d Cir. 2024). A litigant submitting hallucinated citations to a court inherently runs afoul of this requirement because “[a] fake opinion is not ‘existing law’ and citation to a fake opinion does not provide a non-frivolous ground for extending, modifying, or reversing existing law, or for establishing new law.” Id. (quoting Mata v. Avianaca, Inc., 678 F. Supp. 3d 44 (S.D.N.Y. 2023)). A court may “initiate sanctions sua sponte by issuing an order ‘to show cause why conduct specifically described in the order has not violated Rule 11(b).” Kyros L. P.C. v. World Case 3:24-cv-01709-VDO Document 35 Filed 12/15/25 Page 7 of 12
8 Wrestling Ent., Inc., 78 F.4th 532, 543 (2d Cir. 2023) (quoting Fed. R. Civ. P. 11(c)(3)). The power to sua sponte impose Rule 11 sanctions is akin to “a courts inherent power of contempt,” and like contempt, a district court may sua sponte sanction a litigant only upon finding they acted with “subjective bad faith.” In re Pennie & Edmonds LLP, 323 F.3d 86, 90 (2d. Cir. 2003). Subjective bad faith is a “heightened mens rea standard” designed to promote zealous advocacy while deterring improper submissions. Id. at 91. The heightened standard can be satisfied upon finding that a litigant made a “misleading representation to a district court for an improper purpose.” Rankin v. City of Niagara Falls, 293 F.R.D. 375, 387 (W.D.N.Y. 2013), aff’d sub nom, 569 F. App’x 25 (2d Cir. 2014). Accordingly, courts in this circuit have repeatedly found that presenting AI-generated hallucinations as valid caselaw constitutes subjective bad faith. See, e.g., Andre v. Warden, No. 3:24-CV-01295 (VDO), 2025 WL 3281732, at *6 (D. Conn. Nov. 25, 2025). Mata, 678 F. Supp. at 464; Benjamin v. Costco Wholesale Corp., 779 F. Supp. 3d 341, 350 (E.D.N.Y. 2025); Ramirez v. Humala, No. 24-CV-242, 2025 WL 1384161, at *2 (E.D.N.Y. May 13, 2025); Kaur v. Desso, No. 9:25-CV-726, 2025 WL 1895859, at *3 (N.D.N.Y. July 9, 2025) III. DISCUSSION “Artificial intelligence may ultimately prove a helpful tool to assist pro se litigants in bringing meritorious cases to the courts.” Moales v. Land Rover Cherry Hill, No. 3:25-CV-544 (VDO), 2025 WL 1249616, at *3 (D. Conn. Apr. 30, 2025). Yet, repeated misuse of artificial intelligence undermines the genuine role it may ultimately have in contributing to the cause of justice. See id. Accordingly, this Court recently adopted a standard for confronting a party who used artificial intelligence in drafting its pleadings: Case 3:24-cv-01709-VDO Document 35 Filed 12/15/25 Page 8 of 12
9 When a pleading containing fabricated citations or nonexistent law has been identified and a hearing held, this Court shall evaluate two core factors: 1. Verification and Inquiry — Whether counsel or the unrepresented party conducted a reasonable, human-based verification of every cited authority before filing. A reasonable inquiry requires more than reliance on an automated tool; it demands independent confirmation through recognized primary legal sources. The signature of an attorney or the unrepresented party certifies human diligence, not mechanical output. 2. Candor and Correction — Whether counsel or the unrepresented party promptly disclosed the use of AI and corrected the record once inaccuracies were discovered. Candor after filing weighs heavily in mitigation; concealment or minimization aggravates the violation. The duty of candor extends beyond the moment of signature, it continues for as long as the filing remains before the Court. Andre, 2025 WL 3281732, at *6–7 (quoting Mattox v. Prod. Innovations Rsch., LLC, No. 6:24-CV-235, 2025 WL 3012828, at *4 (E.D. Okla. Oct. 22, 2025)). In Andre, this Court also adopted a definition of hallucination in the context of AI-assisted legal research “as one of three errors: (1) fabricated cases (whether a nonexistent case name and citation, an existing case name with an invented citation, or a real citation that leads to a wholly unrelated decision); (2) fabricated quotations from actual cases; and (3) misstatements of law (representations of legal rules, standards, or holdings that are inaccurate, incomplete, or unsupported by any real authority, including AI-generated summaries that distort or materially alter what a case actually decided).” Id. at *6. The third form of hallucination—misstatements of law—is often difficult to discern from instances where a litigant simply misunderstands a case or its holding. In those situations, to determine if a party used AI-generated content without verifying its authenticity, the Court looks to: (1) the prevalence of other obvious or admitted AI-generated content in the litigant’s filings; and (2) the egregiousness of the misstatement. For example, the presence of a Case 3:24-cv-01709-VDO Document 35 Filed 12/15/25 Page 9 of 12
10 fabricated quote in a filing alerts the Court to the use of AI. The Court is then likely to conclude that statements of the law that make little sense or are not tied to any case law, statutes, or rules, are also AI-generated misstatements of the law, rather than misunderstandings by the litigant. On the first Andre factor—verification and inquiry—Plaintiff fails entirely. Under Rule 11(b), a litigant’s signature verifies they conducted a “reasonable inquiry” into the accuracy of their legal arguments and cited authorities. Yet, it is evident that Plaintiff did not read the filings or make any attempt to check whether the assertions, cases and quotations he generated using AI were accurate. In Plaintiff’s Response to the Order to Show Cause, he readily admits to not verifying the contents of his pleadings, stating, “I accepted the output as correct and incorporated it into [my] briefing in good faith, not knowing it was riddled with falsehoods and misrepresentations.” Even a cursory check of Defendant’s briefs would have uncovered a plethora of AI-generated hallucinations, including fake citations, fabricated quotations, and repeated misstatements of law. Nevertheless, in blatant violation of Rule 11(b), Defendant repeatedly signed his filings and accordingly certified to the Court that he was presenting truthful, nonfrivolous arguments. Plaintiff also fails on the second factor, candor and correction. Plaintiff had two opportunities to correct his filings. First, Plaintiff could have admitted the inaccuracies in his Sur-Reply after Defendant uncovered them. Although he confessed to misattributing a quote and misunderstanding some of the case law, he simultaneously continued to cite AI-generated hallucinations—including an entirely fabricated case and cases that have no relationship to the Motion to Dismiss. In that sense, he concealed and minimized the role of AI in his filings while concurrently continuing to violate Rule 11 by presenting new, fictitious citations. Second, Case 3:24-cv-01709-VDO Document 35 Filed 12/15/25 Page 10 of 12
11 Plaintiff had the opportunity to correct his filings when the Court issued its Order to Show Cause. There, he profusely apologized to the Court, yet he failed to identify the precise source of the fictitious content or create a workable list of corrective actions. Plaintiff states that he failed to verify the submissions because he is a pro se plaintiff that lacks legal training. He then proceeds to state that he will “diligently verify[] every legal citation and quotation,” without explaining how he will do so. His statements not only undermine the capabilities of pro se litigants but contradict each other. The Court acknowledges Plaintiff apologized for the submissions, “[b]ut regret and apologies are not necessarily enough to avoid the imposition of sanctions for the submission of non-existent legal authority.” Benjamin, 779 F. Supp. at 350. Accordingly, the Court finds Plaintiff acted in subjective bad faith to violate Rule 11 by repeatedly relying on AI-generated hallucinations and ignoring every opportunity to correct his briefs. He knowingly represented fabricated caselaw to this Court and concealed his actions when given the opportunity to accept responsibility. At the very least, Plaintiff made “misleading representation[s]” to this Court “for an improper purpose.” Rankin, 293 F.R.D. 375 at 387. Namely, he made no attempts to check whether his AI-generated briefs were accurate, and he used these misleading hallucinations to advance his case. Therefore, in line with the approach of other courts in this Circuit, the Court finds that Plaintiff’s actions satisfy the subjective bad faith standard and sua sponte imposes Rule 11 sanctions. The Court turns to fashioning an appropriate sanction. Courts have wide discretion in determining precise form of Rule 11 sanctions to impose. Oliveri v. Thompson, 803 F.2d 1265, 1280 (2d Cir. 1986) (“To deal effectively, fairly, and efficiently with sanction claims, district courts have a difficult task, but wide discretion.”). The sanction imposed “must be limited to what suffices to deter repetition of the conduct or comparable conduct by others similarly Case 3:24-cv-01709-VDO Document 35 Filed 12/15/25 Page 11 of 12
12 situated.” Fed. R. Civ. P. 11(c)(4). In making this determination the Court has a wide variety of sanctions to impose such as “striking the offending paper,” or “issuing an admonition, reprimand, or censure.” Fed. R. Civ. P. 11, advisory committee's note to 1993 Amendments. When warranted, even “dismissal of a complaint is among the permissible sanctions allowed under Rule 11.” Miller v. Bridgeport Bd. of Educ., No. 3:12-CV-01287 (JAM), 2014 WL 3738057, at *10 (D. Conn. July 30, 2014). Here, in the interests of justice and deterrence, the Court ADMONISHES Plaintiff for repeatedly violating Rule 11 by submitting inaccurate, unverified AI-generated content to this Court, and then attempting to mislead the Court by concealing his actions. Additionally, the Court STRIKES Plaintiff’s Response and Sur-Reply from the record, without leave to amend. Finally, the Court cautions Plaintiff that it will impose more stringent Rule 11 sanctions—including dismissal of the case with prejudice—upon identifying AI-generated hallucinations in any future filings. IV. CONCLUSION For the foregoing reasons, the Court ADMONISHES Plaintiff and STRIKES Plaintiffs Response to the Motion to Dismiss (ECF No. 30) and Sur-Reply (ECF No. 32) from the record, without leave to amend. SO ORDERED. Hartford, Connecticut December 15, 2025 /s/Vernon D. Oliver VERNON D. OLIVER United States District Judge Case 3:24-cv-01709-VDO Document 35 Filed 12/15/25 Page 12 of 12
ProvenanceKnow exactly where this document came from.Members see the sourcing behind every authority on DocPost — so you can check the record yourself and cite with confidence.Request access