Four attorneys disqualified and fined by a US district court after submitting AI-fabricated legal authorities
On 8 June, the US District Court for the Northern District of Mississippi issued several sanctions to four US attorneys on grounds of Rule 11 of the Federal Rules of Civil Procedure (Rule 11) and abuse of judicial process, because of inappropriate and unverified AI use in their submissions. The order was brought to the attention of The IPKat by one of the readers. Although the underlying proceedings do not concern an IP dispute, it may be of interest to readers following judicial developments on AI.
Under Rule 11, attorneys have a duty to conduct reasonable inquiry into the law when affixing their signature on documents they submit to courts [see pp. 4-5 of the judgment]. Whether they satisfied their duties in a given case is judged against an objective standard, as opposed to their subjective knowledge or intentions. Moreover, the Court held that misusing AI and submitting fabricated authority constitutes an abuse of judicial process, pointing to a Fifth Circuit’s recent ruling.
Starting with the resident counsel, Ridgeway and McClinton submitted that their involvement with the drafting of the filings was limited and they were both unaware that their co-counsel had used AI. While these arguments were not treated as valid defences allowing the resident attorneys to avoid responsibility for the Rule 11 violation, the Court took into account their honesty, self-reporting to the Mississippi Bar, and sincere apologies when determining their sanctions. As the ‘gateway’ for out-of-state attorneys, resident counsel are responsible for overseeing their submissions, which was a duty both Ridgeway and McClinton failed to fulfil.
Ridgeway and McClinton admitted that they did not review their co-counsel’s drafts or verify whether the cited authorities existed before signing and filing the submissions. Furthermore, McClinton stated that he did not have access to the submission prior to filing, given the implicit permission he had given Williams to do so based on their usual course of practice. However, he accepted that he could have reviewed it upon notification of electronic filing. Consequently, the Court held that neither resident attorney acted in bad faith or purposefully attempted to mislead the Court; nevertheless, they were negligent and careless when affixing their signatures to the relevant court submissions.
Moving on to the out-of-state attorneys, the Court held, in essence, that they knew or reasonably should have known the risks involved in using AI in the legal profession; should, in any case, have read the cases they cite in their submissions (as not doing so is a violation of Rule 11); and acted in bad faith by failing to verify the AI-generated legal authorities.
Wilson stated that she was ‘shocked’ when the Court issued the Show Cause Order and that she was not aware that AI could hallucinate non-existent cases. However, in accordance with the Fifth Circuit’s precedent mentioned above, the Court dismissed this alleged ignorance. Moreover, despite apologising to the Court at the Show Cause hearing, Wilson continued filing AI-generated submissions without verifying the citations in another case two months after the hearing, suggesting her apology was not sincere.
Williams was a more experienced AI user, having installed an AI tool at her firm (where she is a partner) and being involved in the development of the firm’s AI policy. Ironically, she disregarded that policy ‘by blindly relying on’ the said AI tool, demonstrating deliberate bad faith on her part. Additionally, she claimed that, even if the AI-fabricated cases did not exist, the legal propositions she put forward were still accurate and supported by other, existing case law. The Court dismissed this claim, echoing another district court’s ruling that ‘the invitation to consider that actual authorities stand for the proposition that the bogus authorities were offered to support’ must be rejected as this is ‘a stroke of pure luck for … lawyers, and [cannot] remed[y] the waste and harm their misconduct wrought’. One final point to note, which apparently aggravated Williams’s circumstances, was her attempt to evade the Show Cause hearing by providing misleading documentation to the Court to misrepresent her availability.
Against this background, the Court (i) disqualified all four attorneys from participating further in this case, while barring Wilson and Williams from appearing before it for 2 years, (ii) ordered Ridgeway and McClinton to pay $1,000 each, Wilson to pay $2,500, and Williams to pay $3,500 to the registry of the Court, (iii) ordered Wilson to complete an AI training programme with an ethics component to it, and (iv) gave the parties of the original dispute 60 days to find new counsel to represent them.
To be clear, it is not suggested that the use of AI tools in the legal context should be completely banned. As it has been argued elsewhere, AI tools possess many capabilities that can assist lawyers in their daily tasks, legal research processes, and document drafting. Likewise, courts themselves do not impose outright bans on the use of AI.
Considering IP practice, AI tools are being used at numerous stages by various parties, such as trade mark applicants checking for the availability of a sign and attorneys drafting patent claims. What is remarkable is that these uses are no longer considered exceptional: An overview of EUIPO’s stance towards the use of AI demonstrates how integrated the technology has become in, for instance, the trade mark registration process. Similarly, a recent study conducted among the global IP community shows that 85% of the IP professionals surveyed now use AI in some capacity. Moreover, in the UK, Lord Justice Birss has been openly commenting on his use of AI tools, for example, to summarise very long documents to save substantial court time [see here and here for Birss LJ’s comments; also here for uses of AI by US courts].
Although such AI uses are not contested, the increasing risk of over-reliance on these tools brings the issues of hallucination, disclosure of AI use, and assumption of responsibility for AI-generated texts to the fore. Against this backdrop, many cautionary views were expressed, including by Lord Justice Birss, emphasising that it is still the person signing the text who assumes responsibility for its contents, that blaming AI for hallucinated legal authorities cannot serve as a valid excuse to avoid liability, and that there is an increasing need for guidelines allowing not only judges, but also lawyers and experts who submit documents to courts, to assess the appropriateness of their use, disclose it where necessary, and shape their practices accordingly [see here for an example of such guidelines].
It is acceptable, and perhaps even inevitable, to make use of the AI technology when conducting research, say, to review existing trade marks and check the availability of a certain sign for registration; or to identify infringing content on online platforms. Likewise, it can be acceptable to work with AI tools in various stages of drafting takedown notices, patent claims, witness or expert reports demonstrating infringement, or other documents to be submitted to courts. However, lawyers and judges must avoid – and where appropriate report or penalise – blind, over-reliance on the summaries, findings, or suggestions generated by these tools. Signing, and thus, assuming authorship for AI-generated texts without properly reading or sufficiently contributing to them; and failing to disclose AI use are not in line with the abovementioned guidelines and other ‘compliance tips’ adopted, for instance, by the Solicitors Regulation Authority, and, as demonstrated in this case, can have severe consequences.
Background
In a contractual dispute between Tom Withers and the City of Aberdeen, the Court identified a few ‘hallucinatory citations’ found in the submission of both Withers’s and the City’s counsel. Each party had one out-of-state attorney – Kathleen Wilson and Kathryn Williams – and one local, resident counsel – Shauncey Hunter Ridgeway and Mark McClinton – who all stated that they became aware of the said fabricated cases only upon the Court’s Show Cause Order in December 2025.Court’s decision
The Court ruled that all four attorneys had violated Rule 11. However, they had done so to different degrees and therefore received different sanctions.
Under Rule 11, attorneys have a duty to conduct reasonable inquiry into the law when affixing their signature on documents they submit to courts [see pp. 4-5 of the judgment]. Whether they satisfied their duties in a given case is judged against an objective standard, as opposed to their subjective knowledge or intentions. Moreover, the Court held that misusing AI and submitting fabricated authority constitutes an abuse of judicial process, pointing to a Fifth Circuit’s recent ruling.
Starting with the resident counsel, Ridgeway and McClinton submitted that their involvement with the drafting of the filings was limited and they were both unaware that their co-counsel had used AI. While these arguments were not treated as valid defences allowing the resident attorneys to avoid responsibility for the Rule 11 violation, the Court took into account their honesty, self-reporting to the Mississippi Bar, and sincere apologies when determining their sanctions. As the ‘gateway’ for out-of-state attorneys, resident counsel are responsible for overseeing their submissions, which was a duty both Ridgeway and McClinton failed to fulfil.
Ridgeway and McClinton admitted that they did not review their co-counsel’s drafts or verify whether the cited authorities existed before signing and filing the submissions. Furthermore, McClinton stated that he did not have access to the submission prior to filing, given the implicit permission he had given Williams to do so based on their usual course of practice. However, he accepted that he could have reviewed it upon notification of electronic filing. Consequently, the Court held that neither resident attorney acted in bad faith or purposefully attempted to mislead the Court; nevertheless, they were negligent and careless when affixing their signatures to the relevant court submissions.
Moving on to the out-of-state attorneys, the Court held, in essence, that they knew or reasonably should have known the risks involved in using AI in the legal profession; should, in any case, have read the cases they cite in their submissions (as not doing so is a violation of Rule 11); and acted in bad faith by failing to verify the AI-generated legal authorities.
Wilson stated that she was ‘shocked’ when the Court issued the Show Cause Order and that she was not aware that AI could hallucinate non-existent cases. However, in accordance with the Fifth Circuit’s precedent mentioned above, the Court dismissed this alleged ignorance. Moreover, despite apologising to the Court at the Show Cause hearing, Wilson continued filing AI-generated submissions without verifying the citations in another case two months after the hearing, suggesting her apology was not sincere.
Williams was a more experienced AI user, having installed an AI tool at her firm (where she is a partner) and being involved in the development of the firm’s AI policy. Ironically, she disregarded that policy ‘by blindly relying on’ the said AI tool, demonstrating deliberate bad faith on her part. Additionally, she claimed that, even if the AI-fabricated cases did not exist, the legal propositions she put forward were still accurate and supported by other, existing case law. The Court dismissed this claim, echoing another district court’s ruling that ‘the invitation to consider that actual authorities stand for the proposition that the bogus authorities were offered to support’ must be rejected as this is ‘a stroke of pure luck for … lawyers, and [cannot] remed[y] the waste and harm their misconduct wrought’. One final point to note, which apparently aggravated Williams’s circumstances, was her attempt to evade the Show Cause hearing by providing misleading documentation to the Court to misrepresent her availability.
Against this background, the Court (i) disqualified all four attorneys from participating further in this case, while barring Wilson and Williams from appearing before it for 2 years, (ii) ordered Ridgeway and McClinton to pay $1,000 each, Wilson to pay $2,500, and Williams to pay $3,500 to the registry of the Court, (iii) ordered Wilson to complete an AI training programme with an ethics component to it, and (iv) gave the parties of the original dispute 60 days to find new counsel to represent them.
Comment
Barring attorneys from appearing before a court, disqualifying them from proceedings, and imposing non-negligible financial penalties might come to some as a harsh order. However, given the gravity of the issue, this Kat finds all these sanctions reasonable, and would have supported an even harsher order aimed at deterring lawyers from relying blindly [Merpel: Seems quite ironic when lawyers do it, right?] on AI-generated submissions.To be clear, it is not suggested that the use of AI tools in the legal context should be completely banned. As it has been argued elsewhere, AI tools possess many capabilities that can assist lawyers in their daily tasks, legal research processes, and document drafting. Likewise, courts themselves do not impose outright bans on the use of AI.
Considering IP practice, AI tools are being used at numerous stages by various parties, such as trade mark applicants checking for the availability of a sign and attorneys drafting patent claims. What is remarkable is that these uses are no longer considered exceptional: An overview of EUIPO’s stance towards the use of AI demonstrates how integrated the technology has become in, for instance, the trade mark registration process. Similarly, a recent study conducted among the global IP community shows that 85% of the IP professionals surveyed now use AI in some capacity. Moreover, in the UK, Lord Justice Birss has been openly commenting on his use of AI tools, for example, to summarise very long documents to save substantial court time [see here and here for Birss LJ’s comments; also here for uses of AI by US courts].
Although such AI uses are not contested, the increasing risk of over-reliance on these tools brings the issues of hallucination, disclosure of AI use, and assumption of responsibility for AI-generated texts to the fore. Against this backdrop, many cautionary views were expressed, including by Lord Justice Birss, emphasising that it is still the person signing the text who assumes responsibility for its contents, that blaming AI for hallucinated legal authorities cannot serve as a valid excuse to avoid liability, and that there is an increasing need for guidelines allowing not only judges, but also lawyers and experts who submit documents to courts, to assess the appropriateness of their use, disclose it where necessary, and shape their practices accordingly [see here for an example of such guidelines].
It is acceptable, and perhaps even inevitable, to make use of the AI technology when conducting research, say, to review existing trade marks and check the availability of a certain sign for registration; or to identify infringing content on online platforms. Likewise, it can be acceptable to work with AI tools in various stages of drafting takedown notices, patent claims, witness or expert reports demonstrating infringement, or other documents to be submitted to courts. However, lawyers and judges must avoid – and where appropriate report or penalise – blind, over-reliance on the summaries, findings, or suggestions generated by these tools. Signing, and thus, assuming authorship for AI-generated texts without properly reading or sufficiently contributing to them; and failing to disclose AI use are not in line with the abovementioned guidelines and other ‘compliance tips’ adopted, for instance, by the Solicitors Regulation Authority, and, as demonstrated in this case, can have severe consequences.
Image credit: ChatGPT
Four attorneys disqualified and fined by a US district court after submitting AI-fabricated legal authorities
Reviewed by Söğüt Atilla-Aydın
on
Tuesday, June 30, 2026
Rating:
Reviewed by Söğüt Atilla-Aydın
on
Tuesday, June 30, 2026
Rating:
Follow the IPKat on LinkedIn 
No comments:
All comments must be moderated by a member of the IPKat team before they appear on the blog. Comments will not be allowed if the contravene the IPKat policy that readers' comments should not be obscene or defamatory; they should not consist of ad hominem attacks on members of the blog team or other comment-posters and they should make a constructive contribution to the discussion of the post on which they purport to comment.
It is also the IPKat policy that comments should not be made completely anonymously, and users should use a consistent name or pseudonym (which should not itself be defamatory or obscene, or that of another real person), either in the "identity" field, or at the beginning of the comment. Current practice is to, however, allow a limited number of comments that contravene this policy, provided that the comment has a high degree of relevance and the comment chain does not become too difficult to follow.
Learn more here: http://ipkitten.blogspot.com/p/want-to-complain.html