Executive Summary

Generative AI models produce massive amounts of erroneous or fabricated content in legal documents. Lawyers and judicial authorities increasingly rely on AI tools to create legal briefs without verifying hallucinations. An online database maintained by legal scholar Damien Charlotin documented over 1500 cases by June 2026 in which courts sanctioned individuals for AI errors in legal documents – 1000 of these in the United States alone. The actual number of unreported cases is significantly higher, as only discovered cases are recorded.

People

  • Damien Charlotin (Legal scholar, HEC Paris)
  • Deborah Leslie (Deputy District Attorney, Georgia)
  • Hannah Payne (Defendant, Georgia case)

Topics

  • Artificial Intelligence in the Judiciary
  • AI Hallucinations and Data Fabrication
  • Judicial Failure and Lack of Controls
  • Liability and Sanctions

Clarus Lead

The escalation of AI errors in court proceedings reveals a structural failure in the control and verification of critical documents. Particularly alarming is the fact that not only private lawyers, but also judges and prosecutors accept or pass on erroneous texts without verification – with direct consequences for defendants. The case of prosecutor Deborah Leslie demonstrates: even supreme courts are deceived by AI-generated fictions when controls fail. This fundamentally endangers the core principles of legal certainty and justice.

Detailed Summary

Legal scholar Damien Charlotin of HEC Paris maintains a database documenting court proceedings in which AI errors led to sanctions. According to current figures, over 1500 such cases are registered, with the United States significantly overrepresented at approximately 1000 cases. The actual number is considerably higher, as the database only captures cases in which AI hallucinations were actually discovered.

A particularly striking example is the case of Deputy District Attorney Deborah Leslie in Georgia. In May 2026, she filed a motion draft to effect the dismissal of a motion for retrial. The case involved Hannah Payne, who had been sentenced to 13 years imprisonment for murder and false imprisonment. The motion draft contained over 20 references to misquoted or completely fabricated court rulings. Not only did the prosecutor adopt the AI-generated citations without verification, but the judge also copied large sections of the erroneous template into his decision. Result: Leslie was suspended from practicing before the Supreme Court for six months.

Two U.S. federal judges were also censured in 2025 for obvious AI errors in their rulings. However, both showed little insight – one blamed his legal clerk, the other blamed an intern. These reactions indicate a deeper cultural problem: responsibility is delegated rather than verified independently.

Key Points

  • AI systematically produces fabricated quotations and rulings in legal documents that are difficult to detect
  • Over 1500 documented cases worldwide show that lawyers, judges, and prosecutors use AI-generated texts without verification
  • Even the highest courts are deceived by AI hallucinations when controls are lacking
  • The failure of judicial authorities to assume responsibility exacerbates the problem

Critical Questions

  1. Evidence/Data Quality: How comprehensive is Charlotin's database? Which countries and jurisdictions are underrepresented, and could regional differences in data collection explain the high U.S. figures?

  2. Conflicts of Interest: Do AI providers profit financially from the lack of liability of their models? Are there incentives not to document errors?

  3. Causality/Alternatives: Are AI hallucinations the core problem, or is the fault primarily inadequate verification processes? Would stricter client liability solve the problem?

  4. Feasibility/Risks: What technical or procedural solutions could reliably detect AI errors in legal briefs? Would a ban on AI in critical proceedings simply lead to workarounds?

  5. Causality: The Leslie case demonstrates a lack of control at two levels (prosecution + judge). Who bears primary responsibility for these cascading errors?

  6. Feasibility: Can effective liability be established for judges who uncritically adopt AI-generated erroneous citations?

Sources

Primary Source: Justice on the Wrong Path: The Judiciary and the Struggle with AI Garbage – https://www.heise.de/hintergrund/Justitia-auf-Irrwegen-Die-Rechtsprechung-und-der-Kampf-mit-dem-KI-Muell-11313496.html

Verification Status: ✓ June 2026

This text was created with the support of an AI model. Editorial Responsibility: clarus.news | Fact-Check: June 2026