Thomas v. Delaware Technical & Cmty. Coll.

What sanction did the court impose?

Chief Judge Colm F. Connolly denied both motions. The Rule 11 motion was DENIED IN PART AS MOOT (the underlying motions had already been ruled on and denied) and DENIED IN PART WITHOUT PREJUDICE TO RENEW (the fee request was held premature, to be renewed after the merits are resolved when the Court can consider the totality of Thomas's litigation conduct). The motion to strike and bar was DENIED in full, with the Court declining to impose a prefiling bar at that point. The Court flagged its serious concerns on the record: 'Thomas's status as a pro se litigant does not immunize him from Rule 11 sanctions,' citing Karpov v. Karpov, 307 F.R.D. 345, 348 (D. Del. 2015), and 'I am very concerned about Thomas's unverified use of AI. Several of the cases Thomas cites and passages he purports to quote in his filings do not exist.' In place of a prefiling bar or monetary sanctions, the Court imposed a docket-management measure: 'Defendants need not respond to any future filings by Plaintiff unless and until I order a response.'

Why does Thomas v. Delaware Technical & Cmty. Coll. matter for law firms using AI?

The notable doctrinal move in Chief Judge Connolly’s order is structural rather than substantive. Defendants asked the Court for a prefiling bar, and prefiling bars carry their own caselaw: the Third Circuit, like most circuits, requires a finding that less restrictive means would be inadequate before barring a litigant from filing without leave. That analysis is fact-intensive and creates appellate risk. Connolly’s response-control measure achieves much of the same practical effect (defendants are not pulled into responding to filings the Court has not deemed worth a response) without invoking the prefiling-injunction framework. The cost shifts from a substantive limit on Thomas’s right to file to a procedural reordering of the response burden. For partners thinking about how to defend against a serial pro se filer, this is a useful template.

The fee-deferral piece is also worth noting. Connolly held that the Rule 11 fee request was premature ‘at this juncture’ and denied it without prejudice to renew once the merits of the case are resolved and the Court ‘may consider the totality of Thomas’s conduct in litigating the case.’ That framing preserves the defense’s fee option and creates an incentive structure: continued AI-citation conduct after the order will be visible to the Court when the eventual fee motion is filed, so the docket itself becomes the deterrent rather than an immediate sanction. This is consistent with the emerging cost-of-sanctions approach in pro se AI-hallucination cases (see, e.g., Leslie v. IQ Data Int’l Corp., where Magistrate Judge Cannon declined monetary sanctions in favor of dismissal with prejudice for the combined effect of discovery non-compliance and fake citations).

The Karpov v. Karpov citation matters for outside-Delaware practice as well. Several federal courts have noted that pro se status does not immunize from Rule 11, but the D. Del. authority is now anchored to a 2015 in-district decision rather than to the broader Eleventh Circuit or Second Circuit caselaw that has been doing most of the work in the AI-hallucination cases. That gives Delaware practitioners a clean local cite when responding to pro se filings that rely on AI-generated content.