In re Daniel M. Risis

What sanction did the court impose?

On May 9, 2025, Bankruptcy Judge John K. Sherwood issued a stand-alone 'ORDER RE: THE DEBTOR'S USE OF ARTIFICIAL INTELLIGENCE' (ECF No. 132) denying the debtor's representation by 'Kasper' and striking all 16 'Kasper'-signed filings. Key holdings: (1) 'The Bankruptcy Rules do not recognize artificial legal intelligence representatives. Only the Debtor or an attorney (if one is retained) can sign pleadings or present motions on his behalf.' (2) 'The Court has no issue with the Debtor's use of artificial intelligence to aid in the preparation of pleadings. But it is the Debtor's responsibility to sign, proofread and independently verify every aspect of the materials for accuracy pursuant to Rule 9011.' (3) The motion to integrate 'Kasper' with PACER was denied because 'the Debtor has access to PACER like every other party that comes before the Court.' (4) Any future filings the debtor wishes to refile must be signed by Risis himself and must comply with Fed. R. Bankr. P. 9011. The Court cited S.D.N.Y.L.B.R. 9011-1(d) for the AI-use-but-debtor-still-responsible framing. No monetary sanctions were imposed; the relief was procedural (strike + deny without prejudice).

Why does In re Daniel M. Risis matter for law firms using AI?

This case is doctrinally distinct from the rest of the 2025-2026 AI-citation corpus. Most cases involve a pro se litigant or counsel using AI to draft pleadings that contain hallucinated citations; the court then warns or sanctions for the hallucinations. Risis presents the question one step earlier: can a pro se party have an AI agent stand in for them entirely, sign pleadings as their representative, and access the court’s electronic-filing system? Sherwood’s answer is a clean no, anchored in the plain text of Fed. R. Bankr. P. 9011(a).

The bright-line rule (AI as tool, allowed; AI as representative, forbidden) is the most exportable holding in this order. The same rule applies in district-court practice under Fed. R. Civ. P. 11(a), which similarly requires that ‘every pleading, written motion, and other paper must be signed… by a party personally if the party is unrepresented.’ The same exclusion logic applies. For partners thinking about how to advise clients on the limits of AI delegation, In re Risis is the cleanest cite for the proposition that a party cannot outsource its signature obligation to an AI, regardless of how thoroughly the party purports to authorize the delegation in writing.

The PACER-integration denial is the procedural corollary worth flagging. Risis asked the Court to formally integrate ‘Kasper’ with PACER, which would have effectively granted the AI agent filing access. Sherwood denied this as a function of the representation question: an AI agent that cannot sign cannot file. This forecloses the workaround where a party might argue that allowing AI to file is procedurally different from allowing AI to represent. For firms advising clients on AI-tool integrations with court systems, the answer in 2025 Bankr. D.N.J. is that the integration question collapses into the representation question, and the representation question has been answered.

The cite to S.D.N.Y.L.B.R. 9011-1(d) is also worth noting. That local bankruptcy rule is the S.D.N.Y. counterpart to the AI-use-and-verification framing in cases like Schlesinger’s CMSO in Soehlig v. Experian (M.D. Fla.). Bankruptcy courts are quietly building their own AI policies through local rules and individual orders; for practitioners with significant bankruptcy practice, the corpus to monitor is wider than just the headline district-court sanctions cases.