{"slug": "pro-se-litigants-flood-dockets-with-ai-generated-filings", "title": "Pro se Litigants Flood Dockets With AI-Generated Filings", "summary": "A study by MIT researcher Anand Shah and USC researcher Joshua Levy found that 18% of pro se court filings now contain AI-generated text, with the volume of pro se docket entries per court in the first 180 days of a case rising 64% on average since the availability of large language models. The national non-prisoner pro se filing share climbed to 16.8% in fiscal year 2025, up from a historical steady state of roughly 11%. Minnesota federal judge Patrick J. Schiltz called the trend \"an existential threat to the federal courts,\" as pro se plaintiffs lost 96% of their cases from 1998 to 2017.", "body_md": "# Pro se Litigants Flood Dockets With AI-Generated Filings\n\nGizmodo reports a study by **MIT** researcher **Anand Shah** and **USC** researcher **Joshua Levy** finds that since broadly available LLMs appeared, **18%** of pro se filings contain what the authors classify as AI-generated text and the volume of pro se docket entries per court in the first **180 days** of a case rose **64%** on average in the post-AI period. Gizmodo reports the study says national non-prisoner pro se filing share climbed to **16.8%** in **fiscal year 2025**; the study is not yet peer reviewed. Gizmodo also summarizes reporting in **The New York Times**, which interviewed a man who uses AI to generate lawsuits and quoted Minnesota federal judge **Patrick J. Schiltz** calling the phenomenon \"an existential threat to the federal courts.\"\n\n### What happened\n\nGizmodo reports a study by **MIT** researcher **Anand Shah** and **USC** researcher **Joshua Levy** that finds **18%** of pro se (self-represented) filings include text the authors classify as AI-generated, and that the total volume of pro se docket entries per court in the first **180 days** of a case increased **64%** on average across the post-AI period. Gizmodo reports the study also finds national non-prisoner pro se filing share rose from an approximately **11%** historical steady state to **16.8%** in **fiscal year 2025**; the study has not yet been peer reviewed. Gizmodo summarizes additional coverage in **The New York Times**, which interviewed an individual who uses AI to draft lawsuits and quoted Minnesota federal judge **Patrick J. Schiltz** calling the trend \"an existential threat to the federal courts.\" The Times is also cited for the long-run statistic that pro se plaintiffs lost **96%** of their cases from 1998-2017.\n\n### Editorial analysis - technical context\n\nWidely available large language models (LLMs) lower the time and skill required to draft legal pleadings, which makes the production of mass, low-quality complaints feasible. Industry-pattern observations show automated drafting increases the rate of hallucinated or legally baseless claims because current LLM outputs can mix plausible legal language with incorrect facts or citations. Detection is technically challenging: stylistic analysis and heuristics can flag likely AI-generated text but produce false positives, and metadata-based provenance is uneven across filings and platforms.\n\n### Industry context\n\nCourts and clerks traditionally triage pro se filings that historically originated largely from incarcerated litigants; Gizmodo reports the new rise is concentrated among non-prisoner filers per the study. Observed patterns in comparable administrative systems suggest a surge of low-quality submissions raises processing costs, lengthens backlogs, and prompts debates about access to justice versus administrative gatekeeping. Public reporting highlights that some judges view the influx as a systemic burden; others and defenders of pro se access emphasize that litigation remains one of the few remedies for underserved claimants.\n\n### What to watch\n\nFor observers and practitioners, useful indicators include changes in courts' filing rules, new guidance from judicial councils on AI-authored documents, adoption of e-filing provenance or attestation requirements, and pilot deployments of automated triage and detection tools. Also monitor follow-up academic work and peer review of the Shah and Levy study, and whether administrative data show sustained changes to pro se filing success rates or case-processing times.\n\n### Practical takeaway\n\nThe episode illustrates a concrete misuse vector for LLMs with operational consequences for legal systems and a policy tradeoff between reducing frivolous filings and preserving low-cost access to courts.\n\n## Scoring Rationale\n\nThis story documents a concrete, measurable misuse of LLMs that affects legal workflows, detection needs, and policy debate-relevant for practitioners tracking operational and governance risks. It is notable but not a frontier-model release.\n\nPractice interview problems based on real data\n\n1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.\n\n[Try 250 free problems](/problems)", "url": "https://wpnews.pro/news/pro-se-litigants-flood-dockets-with-ai-generated-filings", "canonical_source": "https://letsdatascience.com/news/pro-se-litigants-flood-dockets-with-ai-generated-filings-96cb5eb2", "published_at": "2026-05-26 11:13:27.325022+00:00", "updated_at": "2026-05-26 11:13:30.566058+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "generative-ai", "ai-policy", "ai-ethics"], "entities": ["MIT", "USC", "Anand Shah", "Joshua Levy", "Gizmodo", "The New York Times", "Patrick J. Schiltz"], "alternates": {"html": "https://wpnews.pro/news/pro-se-litigants-flood-dockets-with-ai-generated-filings", "markdown": "https://wpnews.pro/news/pro-se-litigants-flood-dockets-with-ai-generated-filings.md", "text": "https://wpnews.pro/news/pro-se-litigants-flood-dockets-with-ai-generated-filings.txt", "jsonld": "https://wpnews.pro/news/pro-se-litigants-flood-dockets-with-ai-generated-filings.jsonld"}}