October 18, 2026 (in 16 days): Tenth Circuit: comments close on 2027 rules, incl. proposed Rule 46.5(B)(4) generative-AI review →

Carbonell v. United States

U.S. District Court, Western District of Texas, Austin Division · W.D. Tex. · Texas bar guidance

Conduct

The government told the court that a § 2255 movant's objections contained multiple 'hallucinated' case citations and asked for sanctions.

Consequence

Here the court denied the § 2255 motion (waiver and merits) and, in a one-line footnote, declined to impose any sanctions.

Lesson

In this case an opposing party's request for sanctions over apparent AI citations did not produce any finding or sanction.

Other

Verified September 29, 2026

Citation
Carbonell v. United States, No. 1:25-CV-637-RP, [1:23-CR-99-RP], Order (W.D. Tex. June 30, 2026) (Pitman, J.), ECF No. 78 (criminal docket)
Decided
June 30, 2026

Summary

Rene Carbonell moved under 28 U.S.C. § 2255 to vacate his federal sentence, and Magistrate Judge Dustin Howell recommended denial, finding six of thirteen grounds barred by the appeal and collateral attack waiver in Carbonell's plea agreement and the ineffective assistance claims without merit. Carbonell filed objections (Dkt. 75). As the district court's order recounts, the United States filed a limited response "noting that Carbonell's Objections contain multiple 'hallucinated' case citations seemingly created by the improper use of generative artificial intelligence." According to a footnote in the order, the government "strongly encouraged" the court to "consider appropriate sanctions against Movant for his repeated use of fictitious and/or misleading citations in documents filed with this Court." Carbonell filed a reply. The order does not identify the challenged citations or say whether the court independently confirmed that they were fictitious.

AI tool:
Unidentified (the order recounts the government's assertion that the objections contain 'hallucinated' citations 'seemingly created by the improper use of generative artificial intelligence'; the court makes no finding)
This case summary is informational only. Verify the underlying opinion or order against the primary source before relying on it in any filing or client matter.

How did the court address the AI-related conduct?

Judge Robert Pitman reviewed the report and recommendation de novo, overruled Carbonell's objections, and adopted the recommendation as the court's order, denying the § 2255 motion. Treating the objection to the magistrate judge's denial of an evidentiary hearing as an appeal of a magistrate judge's order, the court found that denial not clearly erroneous or contrary to law and affirmed it. A certificate of appealability was denied. On the government's sanctions request, the court's footnote responds in one sentence: "The Court declines to impose any sanctions." No reason is given, and the order does not address whether the challenged citations were in fact fictitious or how they came to be in the objections. The adopted recommendation rested on the plea agreement's waiver of collateral attack for six of the thirteen grounds and on the lack of merit in the ineffective-assistance claims.

Why does Carbonell v. United States matter for law firms using AI?

Carbonell is a short adopting order in a federal post-conviction case, and the AI issue appears only in the procedural history and a footnote. The United States, responding to the movant’s objections, told the court that the objections contained multiple “hallucinated” case citations that seemed to have been created by improper use of generative AI, and it encouraged the court to consider sanctions for “repeated use of fictitious and/or misleading citations.”

The court did not engage with that request beyond a single sentence declining it. It resolved the objections by adopting the magistrate judge’s analysis that six of thirteen grounds were waived by the plea agreement and that the ineffective-assistance claims failed, and it affirmed the denial of an evidentiary hearing under the deferential clear-error standard.

The entry records a court declining sanctions after an opposing party raised apparent AI citations. The order does not say why, and it does not confirm or reject the government’s characterization of the citations.

Implications for your firm

Operational steps a firm reading this case may wish to consider documenting. Strategic and rule-application calls belong to your firm's attorneys.

  • Document the specific citations when asking a court to sanction an opponent's AI-generated filings. Here the government's request is recounted only in general terms, and the court declined sanctions without discussion.
  • Review post-conviction filings for citation accuracy even when the merits are decided on other grounds. The court resolved the motion on the report and recommendation's waiver and ineffective-assistance analysis.

Sources

Primary sources

Unverified claims:
  • AI attribution: the only AI reference is the court's summary of the government's response; the court makes no finding that Carbonell used AI or that the citations were fictitious. The tracker's AI coding rests on the government's allegation.
  • Whether Carbonell filed his objections pro se is not stated in this order or in the report and recommendation (ECF No. 72); pro_se is left false pending confirmation from the docket's party listing.