Abstract
AI-generated summaries may soon have a clearer path into the courtroom. The December 1, 2024 amendment to Federal Rule of Evidence 1006 clarified that summaries of voluminous records can be admitted as substantive evidence, provided the underlying evidence is properly handled. This note breaks down what this means, why it matters, and the safeguards that will still be required.
The problem: too many documents to review
Lawyers regularly deal with thousands of pages of financial records, rent ledgers, invoices, emails, medical records and other documents that cannot realistically be reviewed one by one during a trial.
What Rule 1006 allows
Federal Rule of Evidence 1006 provides a solution. It allows a party to present a summary, chart or calculation to prove the content of voluminous admissible records that cannot conveniently be examined in court.[1]
Since December 1, 2024, the rule expressly states that these summaries may be admitted as evidence, even when the underlying documents have not themselves been introduced into evidence.[1]
This clarification matters because a Rule 1006 summary is not merely a visual aid. When the proper foundation is established, the summary becomes substantive evidence that the judge or jury may rely on, a principle that federal courts of appeals had already recognized well before the amendment, which codified and clarified this line of case law rather than creating it.[2][3][4] The summary nonetheless remains subject to the Rule 403 balancing test, and may still be excluded where its probative value is substantially outweighed by a danger of unfair prejudice or of misleading the jury.[1]
Where artificial intelligence fits
This creates an interesting path for the use of artificial intelligence.
An AI system could help extract information from thousands of documents, identify transactions, organize dates, calculate totals and produce a clear summary of the underlying evidence.
But the fact that AI helped create the summary would not automatically make it admissible.
The safeguards that still apply
The underlying records must still be admissible. They must be made available to the opposing party, and the final summary must fairly and accurately reflect their contents.[1][3]
The party offering the summary would also need to explain how it was created. This may require preserving the documents provided to the AI system, the prompts used, the model version, the generated output and the human verification process.[5]
Most importantly, a lawyer or qualified witness should independently verify every figure and adopt the final summary as accurate. The safest approach is therefore not to present an unexplained AI output as evidence, but to present a human-reviewed Rule 1006 summary created with the assistance of AI.
A proposed Rule 707: revised and under further study, not adopted
A proposed Federal Rule of Evidence 707 could eventually introduce more specific reliability requirements for machine-generated evidence offered without an expert witness. After a public comment period that ran from August 15, 2025 to February 16, 2026 and drew more than 70 written comments, the Advisory Committee on Evidence Rules decided at its May 7, 2026 meeting not to recommend action on the proposal at that time. As stated in its May 17, 2026 report to the Standing Committee, the Committee instead revised the draft rule and plans to have it vetted by technology and AI-law experts at its Fall 2026 meeting, together with the separate question of “deepfake” evidence under a proposed Rule 901(c).[6][7]
Even if a version of Rule 707 were eventually approved, it would not take effect before December 2028 at the earliest. For the foreseeable future, the existing rules therefore provide the operative framework.
Existing rules, carefully applied
AI can assist with the preparation of the evidence, but admissibility will continue to depend on the reliability of the process, the admissibility of the underlying records and the ability of a human witness to explain and verify the final result.
The future of AI-generated evidence may therefore arrive not through a completely new evidentiary system, but through the careful application of rules that already exist.