Should a pro se litigant be sanctioned for (likely) using AI in his appellate brief which hallucinated nonexistent case quotes? CA7: No, but we have some choice words.
This is how power hides.
Jones v. Kankakee County Sheriff's Department - CA7
This comes from a short (7 page) and otherwise uneventful opinion which considers review of a district court's denial of a recusal motion and Younger abstention.
What I found interesting, however, is the final section quoted in part below which addresses the (likely) use of AI by the pro se litigant which hallucinated quotations that appear nowhere in the cases that they are attributed to. SCUDDER writes, with whom BRENNAN and ST. EVE join:
We close with a few words on non-existent quotations Jones attributed to cases he relied on in his appellate brief. To our eye, the error has all the hallmarks of a so-called AI “hallucination,” a circumstance where an AI large language model generates an output that is fictional, inaccurate, or nonsensical. News accounts over the last few years leave no doubt that the consequences of AI hallucinations can be very serious and worrisome. Equally clear is the enormous investment of human and financial capital to enhance the accuracy of AI generally and LLM output.
[...]
In response to a show cause order, Jones insists that he did not use AI to prepare his brief and, even more, that “[m]isattributed quotations and incorrect citations happen all of the time.” At one level, Jones’s observation is fair, for we often see inaccurate legal representations from pro se litigants. And most of the time, absent an indication of knowledge of falsity or an intent to mislead, we move past the misstatements and resolve the appeals, mindful that pro se litigants al- most always lack legal training. Approximately 60% of our caseload in recent years includes at least one party appearing pro se. We have learned how to resolve those cases with the care all litigants deserve without getting bogged down in unwitting misstatements or untidy filings. In our experience, pro se litigants do their best with the resources available to them.
At another level, however, we doubt Jones’s representation and continue to believe he used a generative AI application to prepare his brief. His brief is meticulous in its presentation—very cleanly and professionally formatted, employing prose and citation formats we rarely see from pro se litigants, and, above all else, attributing non-existent quotations to real cases falling within the area of law implicated by the issues Jones presents on appeal.
Whether we are right or Jones is right need not detain us, for we have no reason to believe his misstatements of law were knowing or intentional. Nor do we have any reason to believe he has the training necessary or resources available to check the accuracy of legal citations. In the circumstances before us, then, and mindful that before today’s decision we have not supplied any guidance on the use of AI by pro se litigants, we stop short of imposing any form of sanction on Jones.
[...]
As pro se litigants employ AI to assist with court filings, a basic reminder seems wise. Accuracy and honesty matter. Indeed, the submission of a legal filing constitutes a representation to a court and, as the Federal Rules of Civil Procedure put the point, both attorneys and unrepresented parties are certifying …
This is how power hides.
Jones v. Kankakee County Sheriff's Department - CA7
This comes from a short (7 page) and otherwise uneventful opinion which considers review of a district court's denial of a recusal motion and Younger abstention.
What I found interesting, however, is the final section quoted in part below which addresses the (likely) use of AI by the pro se litigant which hallucinated quotations that appear nowhere in the cases that they are attributed to. SCUDDER writes, with whom BRENNAN and ST. EVE join:
We close with a few words on non-existent quotations Jones attributed to cases he relied on in his appellate brief. To our eye, the error has all the hallmarks of a so-called AI “hallucination,” a circumstance where an AI large language model generates an output that is fictional, inaccurate, or nonsensical. News accounts over the last few years leave no doubt that the consequences of AI hallucinations can be very serious and worrisome. Equally clear is the enormous investment of human and financial capital to enhance the accuracy of AI generally and LLM output.
[...]
In response to a show cause order, Jones insists that he did not use AI to prepare his brief and, even more, that “[m]isattributed quotations and incorrect citations happen all of the time.” At one level, Jones’s observation is fair, for we often see inaccurate legal representations from pro se litigants. And most of the time, absent an indication of knowledge of falsity or an intent to mislead, we move past the misstatements and resolve the appeals, mindful that pro se litigants al- most always lack legal training. Approximately 60% of our caseload in recent years includes at least one party appearing pro se. We have learned how to resolve those cases with the care all litigants deserve without getting bogged down in unwitting misstatements or untidy filings. In our experience, pro se litigants do their best with the resources available to them.
At another level, however, we doubt Jones’s representation and continue to believe he used a generative AI application to prepare his brief. His brief is meticulous in its presentation—very cleanly and professionally formatted, employing prose and citation formats we rarely see from pro se litigants, and, above all else, attributing non-existent quotations to real cases falling within the area of law implicated by the issues Jones presents on appeal.
Whether we are right or Jones is right need not detain us, for we have no reason to believe his misstatements of law were knowing or intentional. Nor do we have any reason to believe he has the training necessary or resources available to check the accuracy of legal citations. In the circumstances before us, then, and mindful that before today’s decision we have not supplied any guidance on the use of AI by pro se litigants, we stop short of imposing any form of sanction on Jones.
[...]
As pro se litigants employ AI to assist with court filings, a basic reminder seems wise. Accuracy and honesty matter. Indeed, the submission of a legal filing constitutes a representation to a court and, as the Federal Rules of Civil Procedure put the point, both attorneys and unrepresented parties are certifying …
Should a pro se litigant be sanctioned for (likely) using AI in his appellate brief which hallucinated nonexistent case quotes? CA7: No, but we have some choice words.
This is how power hides.
Jones v. Kankakee County Sheriff's Department - CA7
This comes from a short (7 page) and otherwise uneventful opinion which considers review of a district court's denial of a recusal motion and Younger abstention.
What I found interesting, however, is the final section quoted in part below which addresses the (likely) use of AI by the pro se litigant which hallucinated quotations that appear nowhere in the cases that they are attributed to. SCUDDER writes, with whom BRENNAN and ST. EVE join:
We close with a few words on non-existent quotations Jones attributed to cases he relied on in his appellate brief. To our eye, the error has all the hallmarks of a so-called AI “hallucination,” a circumstance where an AI large language model generates an output that is fictional, inaccurate, or nonsensical. News accounts over the last few years leave no doubt that the consequences of AI hallucinations can be very serious and worrisome. Equally clear is the enormous investment of human and financial capital to enhance the accuracy of AI generally and LLM output.
[...]
In response to a show cause order, Jones insists that he did not use AI to prepare his brief and, even more, that “[m]isattributed quotations and incorrect citations happen all of the time.” At one level, Jones’s observation is fair, for we often see inaccurate legal representations from pro se litigants. And most of the time, absent an indication of knowledge of falsity or an intent to mislead, we move past the misstatements and resolve the appeals, mindful that pro se litigants al- most always lack legal training. Approximately 60% of our caseload in recent years includes at least one party appearing pro se. We have learned how to resolve those cases with the care all litigants deserve without getting bogged down in unwitting misstatements or untidy filings. In our experience, pro se litigants do their best with the resources available to them.
At another level, however, we doubt Jones’s representation and continue to believe he used a generative AI application to prepare his brief. His brief is meticulous in its presentation—very cleanly and professionally formatted, employing prose and citation formats we rarely see from pro se litigants, and, above all else, attributing non-existent quotations to real cases falling within the area of law implicated by the issues Jones presents on appeal.
Whether we are right or Jones is right need not detain us, for we have no reason to believe his misstatements of law were knowing or intentional. Nor do we have any reason to believe he has the training necessary or resources available to check the accuracy of legal citations. In the circumstances before us, then, and mindful that before today’s decision we have not supplied any guidance on the use of AI by pro se litigants, we stop short of imposing any form of sanction on Jones.
[...]
As pro se litigants employ AI to assist with court filings, a basic reminder seems wise. Accuracy and honesty matter. Indeed, the submission of a legal filing constitutes a representation to a court and, as the Federal Rules of Civil Procedure put the point, both attorneys and unrepresented parties are certifying …
0 Comments
0 Shares
239 Views
0 Reviews
