AI use in the courts

In Monitoring No.5 we have already considered the use of AI in law enforcement activity. Meanwhile, a new trend is taking shape, that is, limiting the use of AI in litigation. Courts began receiving complaints about AI-generated attorney opinions or party arguments containing references to non-existent court cases, rules of law, or unsupported arguments.

For example, in 2023, the New York District Court imposed a $5000 fine on parties and their representatives for filing ChatGPT- generated written representations that included quotations from at least 6 non-existent court¹ decisions. Another example is the decision of the Tax Chamber Tribunal in the UK 2023: in a tax dispute, the defense used at least 4 references to IRS investigations that did not² exist.

Therefore, in July 2024, the U.S. District Court for the Western District of North³ Carolina , restricted the use of AI to help⁴ attorneys form opinions. It is stated that factual and legal citations in court documents prepared using AI must be verified by the filing parties. Any document submitted to the Court must be accompanied by a stipulation that:

- No AI was used in the preparation of the document, except for AI embedded in standard online sources of legal databases (as Westlaw, Lex1s).

- Every statement and reference to sources in the document should be checked for accuracy.

Similar guidelines have been adopted by⁵ other courts in the US, e.g. Manitoba,⁶

Yukon, etc., as well as by courts in a number of other countries, e.g. UK (on the use of chatbots), New Zealand,⁷

Australia, etc.

It is being established that when using AI tools in litigation parties and representatives must:

− Understand how the tools work. For example, the quality of the generated legal response depends on the data for chatbot training, as well as the quality of the user's request.

− Comply with privacy rules: any personal information entered into the chatbot is stored and can be used in requests from other users.

− If parties are representing themselves in court on their own (without lawyers) - notify the court of the use of AI in documents, so that participants are aware of the risks.

- Verify the accuracy of information if AI was used to generate the documents.

The text generated by AI, shall be verified, so that:

1) A version of AI that was trained on obsolete data that does not incorporate more recent case law or legislative changes, was not used.

2) The generated information was complete and accurate (did not refer to fictitious court cases or rules).

3) The practices used were applicable to the particular jurisdiction and were not taken from other jurisdictions that have different substantive laws and procedural requirements.

Russia’s experience

In July 2024, the website of the Supreme Court of the Russian Federation published its statement on the possibilities to use AI in⁸ preparation for court action. Moreover, the idea of connecting AI to the State Automated System “Justice” is under consideration.

Nevertheless, it is recommended that the Supreme Court develop requirements for the use of AI in Russian courts, including:

1) Notification of the court and participants in the trial if AI was used to prepare documents during the trial.

2) The requirement to verify the accuracy of all references to legislation, jurisprudence and other sources.

From the monitoring issue No. 7, July 2024. Download the full issue (PDF) · issue page at the Gaidar Institute