Tenth Circuit Warning: AI-Fabricated Case Citations Are Sanctionable; Extortion-Driven Threats Usually Fail the Asylum “Nexus” Requirement Absent Evidence of Protected-Ground Animus

1. Introduction

Picon-Diaz v. Bondi (10th Cir. Feb. 13, 2026) is a petition for review of a Board of Immigration Appeals (“BIA”) decision affirming an Immigration Judge’s (“IJ”) denial of asylum and withholding of removal to a Colombian family of five: Martha Liliana Picon-Diaz, her husband Hernando Jose Anaya-Causil, and their three minor children.

The family entered the United States in 2023 and was charged as removable under 8 U.S.C. § 1182(a)(6)(A)(i). Proceeding pro se before the IJ, they described fears stemming from: (1) criminal extortion and violence against Hernando’s uncle (shot after stopping extortion payments), (2) threats by neighborhood youths against their oldest child, (3) kidnapping and robbery of Martha’s brother, and (4) generalized country conditions involving guerrilla recruitment and cartel activity.

The central legal issues were: (a) whether the family proved persecution (past or future) “on account of” a protected ground—especially “membership in a particular social group” based on family ties—and (b) whether certain arguments were barred for failure to exhaust before the BIA under 8 U.S.C. § 1252(d)(1). In a notable ancillary holding, the panel issued a pointed warning to counsel for citing an apparently fabricated Tenth Circuit decision—likely the product of careless use of generative AI.

2. Summary of the Opinion

The Tenth Circuit denied the petition for review. It upheld the agency’s determination that Petitioners failed to establish the required nexus between feared harm and a protected ground. Even assuming the proposed “family” particular social group was cognizable, the record did not compel a finding that family membership was “one central reason” for any feared persecution, as opposed to criminals’ financial motives and generalized criminality.

The court also held Petitioners’ argument that the IJ failed to develop the record was unexhausted because it was not presented to the BIA, and thus the court declined to consider it.

Finally, the court warned Petitioners’ counsel about relying on a case captioned “Rodriguez-Romero v. Garland, 60 F.4th 1283 (10th Cir. 2023),” which the panel determined does not exist. The court declined to impose sanctions in this case, but emphasized that using fabricated authorities is sanctionable and undermines the adversary system.

3. Analysis

3.1. Precedents Cited

(a) Framing the scope of review and standards

  • Neri-Garcia v. Holder, 696 F.3d 1003 (10th Cir. 2012): The panel reiterated that when the BIA issues a brief order, the court reviews the BIA’s decision but may consult the IJ’s fuller reasoning to understand the grounds the BIA adopted. This mattered because the nexus analysis was developed more fully by the IJ, then endorsed by the BIA.
  • Miguel-Pena v. Garland, 94 F.4th 1145 (10th Cir. 2024): The court drew from Miguel-Pena for (i) de novo review of legal conclusions, (ii) substantial-evidence review of factual findings such as nexus, and (iii) the nature of issue exhaustion under 8 U.S.C. § 1252(d)(1) as a mandatory claim-processing rule. Miguel-Pena supplied the doctrinal pathway for denying review of the unexhausted “IJ failed to develop the record” argument.

(b) The asylum/withholding burden and the “nexus” requirement

  • Dallakoti v. Holder, 619 F.3d 1264 (10th Cir. 2010): Used for the baseline proposition that asylum requires a nexus to a protected ground and that withholding has a higher burden. Dallakoti anchored the opinion’s conclusion that failing asylum’s nexus necessarily defeats withholding as well.
  • Orellana-Recinos v. Garland, 993 F.3d 851 (10th Cir. 2021): This was the opinion’s key analogue. There, threats to a mother were contingent on a gang’s recruitment aims toward her son. The court assumed a family-based social group but still found no nexus because the persecutor’s real motive was recruitment, not animus toward the family. Picon-Diaz applies the same logic: extortionists’ threats toward “family” are treated as instrumental pressure to secure payment, not persecution “on account of” family membership.

(c) AI hallucinations, fabricated authorities, and sanctions warnings

  • Wadsworth v. Walmart Inc., 348 F.R.D. 489 (D. Wyo. 2025): Cited for the proposition that it is well known AI tools can generate fake cases, and for the definition of “AI hallucinations” as fabricated sources generated by an AI model. The panel used Wadsworth to explain the likely mechanism of counsel’s error.
  • Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023): Cited to underscore that careless AI use wastes resources and can damage the legal system’s credibility—factual context supporting a warning.
  • Park v. Kim, 91 F.4th 610 (2d Cir. 2024): Quoted for the principle that relying on fake opinions to persuade a court is an “abuse of the adversary system,” and noted as sanctionable conduct. Park supplied appellate-level reinforcement (not just district court admonitions) for the seriousness of fabricated citations.
  • Grant v. City of Long Beach, 96 F.4th 1255 (9th Cir. 2024): Referenced as another appellate example recognizing sanctionability, reinforcing that this is not an isolated or merely local concern.

(d) The “nonexistent” case and the panel’s verification method

  • United States v. Wesley, 60 F.4th 1277 (10th Cir. 2023) and United States v. Diaz-Menera, 60 F.4th 1289 (10th Cir. 2023): The panel explained that the Federal Reporter citation offered for the purported immigration case actually pointed to unrelated criminal cases. This discussion is significant because it models a concrete judicial cross-check: caption mismatch + reporter pagination mismatch + quotation mismatch.
  • Romero v. Garland, No. 21-9515, 2021 WL 6061846 (10th Cir. Dec. 20, 2021): The court noted a superficially similar caption (petitioner name includes “Rodriguez Romero”), but clarified it did not contain the purported quotations and was irrelevant to counsel’s arguments—further confirming the “Rodriguez-Romero v. Garland” citation was fabricated.

3.2. Legal Reasoning

(a) The nexus inquiry dominated the merits

The court followed the statutory framework: asylum requires showing persecution or a well-founded fear of persecution “on account of” a protected ground (8 U.S.C. § 1101(a)(42)), and the protected ground must be “at least one central reason” for the harm (8 U.S.C. § 1158(b)(1)(B)(i)). Building on Orellana-Recinos v. Garland, the panel treated extortion threats as typically motivated by financial gain; family members may be threatened as leverage, but leverage is not the same as persecution because of family status.

Critically, the court applied substantial-evidence review: it did not ask whether a nexus theory was possible, but whether the record compelled a contrary finding (8 U.S.C. § 1252(b)(4)(B)). The panel found it did not, emphasizing:

  • Petitioners did not actually testify to a causal link between the uncle’s 2020 shooting and the son’s 2022 threats; thus, the “temporal and familial link” was argued on appeal but not supported by record evidence.
  • The extortion-related threats were contingent on nonpayment; the uncle continued paying; and Petitioners had not been directly harmed or threatened (aside from the unrelated “Tito” incident). That contingency supported the inference that the motive was payment, not hostility to the family “per se.”
  • Petitioners’ “likely motivation” theory—accepted by the court as a kind of circumstantial approach in principle—failed on this record because the inferences did not rise to the level needed to overturn the agency under substantial-evidence review.

(b) Particular social group: acknowledged, but not outcome-determinative

Petitioners argued the agency failed to recognize a family-based particular social group (“members of a family that had already been targeted by criminal organizations”). The panel responded that, whatever the IJ’s phrasing (“could not decipher”), the BIA did address Petitioners’ asserted group as “the ‘family of a primary target of a criminal organization.’” The court then treated the case as turning on nexus rather than definitively resolving cognizability, noting the government’s contention that the BIA did not make a cognizability finding but that the court “need not decide” because the BIA effectively assumed it for the nexus analysis.

(c) Issue exhaustion barred the “IJ failed to develop the record” argument

Petitioners’ appellate claim that the IJ failed to develop the record—by not eliciting testimony supporting family membership, resistance to extortion, or imputed political opinion—was rejected as unexhausted. Applying Miguel-Pena v. Garland, the panel required Petitioners to have presented “the same specific legal theory” to the BIA. Their BIA briefing argued instead that the IJ should have inferred a particular social group from the facts, not that the IJ failed a duty to develop the record. Because the government timely invoked exhaustion, the court enforced it and declined review.

(d) Fabricated citations: warning without sanctions

The opinion’s most forward-looking move is institutional: it identifies an apparently fabricated precedent, attributes it to likely generative-AI misuse, and flags sanction authority (Fed. R. App. P. 38; 10th Cir. R. 46.5(B), (C)) while declining to sanction in this case. By citing Mata v. Avianca, Inc. and Park v. Kim, the panel places AI hallucinations within established categories of sanctionable litigation abuse: wasting resources, misleading the tribunal, and undermining adversarial integrity.

3.3. Impact

(a) Immigration law: reinforcing limits of family-based nexus in extortion settings

Although designated as “not binding precedent” (an “Order and Judgment”), the decision is a clear signal of how the Tenth Circuit continues to analyze family-based theories where the alleged persecutor’s primary motive is extortion or other instrumental criminal objectives. Building on Orellana-Recinos v. Garland, the opinion reinforces a practical rule: threats to family members used as leverage to secure payment are commonly treated as financially motivated, with family status incidental—unless record evidence shows independent animus toward the family.

For future applicants, the case underscores the evidentiary burden on nexus: it is not enough to show that criminals threatened family members; applicants must develop record proof tying threats to the protected ground as a central reason (statements, patterns, targeting across time/contexts, or continuing pursuit even when the instrumental aim is satisfied).

(b) Appellate practice: a sharpened judicial response to AI-generated false authorities

The warning section is likely to be cited (for persuasive value) in disputes about attorney competence and sanctions where fabricated citations appear. The panel’s approach is notable in three ways:

  • Verification transparency: it documents how the court confirmed the citation’s falsity (reporter cross-checking and quotation mismatch).
  • Norm-setting: it states that AI use is not “inherently problematic,” but “careless use” that yields fake authorities is unacceptable.
  • Sanctions roadmap: it cites multiple appellate and district sanction decisions, plus rule-based sanction authority, while issuing a cautionary “last chance” warning.

Practically, the opinion incentivizes more robust citation-checking practices—particularly when briefs rely on AI-assisted research or drafting.

4. Complex Concepts Simplified

  • Protected grounds: The five statutory reasons asylum can be granted: race, religion, nationality, particular social group, political opinion.
  • Particular social group (PSG): A legally recognized “group” (often including family units) that can qualify as a protected ground. Even if a PSG is valid, the applicant must still prove the harm is because of membership in that PSG.
  • Nexus / “on account of”: The causal link between the persecutor’s motive and a protected ground. If the persecutor’s real reason is money (extortion) or recruitment, nexus usually fails unless family status (or another ground) is a central reason.
  • “One central reason”: The protected ground must be an important motive for persecution, not minor or incidental. The court uses the shorthand from Orellana-Recinos v. Garland: not “incidental, tangential, superficial, or subordinate.”
  • Substantial-evidence review: The court defers to agency fact findings unless the record would compel a different conclusion. This is a high bar for petitioners on appeal.
  • Issue exhaustion: You generally cannot raise a new legal theory in the court of appeals if you did not first present it to the BIA. In this case, arguing “the IJ failed to develop the record” was treated as different from arguing “the IJ should have inferred our PSG.”
  • AI hallucination (legal context): When an AI tool generates fake cases/quotes/citations. Courts treat filing such fabrications as potentially sanctionable misconduct, even if the fabrication was unintentional.

5. Conclusion

Picon-Diaz v. Bondi does two important things. On the merits, it reaffirms that extortion-driven threats—even when aimed at family members—often fail the asylum nexus requirement where the record supports the inference that criminals are pursuing financial gain rather than animus toward a protected group, consistent with Orellana-Recinos v. Garland. Procedurally, it applies strict issue exhaustion under 8 U.S.C. § 1252(d)(1).

Institutionally, the opinion’s most consequential message may be its warning on fabricated authorities: the Tenth Circuit signals heightened scrutiny of citations, frames AI-generated fake cases as a known risk, and places counsel on notice that such conduct is sanctionable—even when linked to careless AI use.