Federal Rule 707 Proposal Sets AI-Generated Evidence Standards
The Advisory Committee proposes Federal Rule 707 for admitting AI-generated evidence without experts.
Why it matters: AI-generated evidence is increasingly common in litigation, requiring clear court standards. This rule will guide lawyers and judges on reliability and admissibility of such digital evidence.
- Rule 707 allows machine-generated evidence admission using Rule 702 reliability standards without expert witnesses.
- Public comments closed February 16, 2026; committee voted in May 2026 on the proposal.
- Rule targets AI evidence but excludes undefined 'simple scientific instruments,' creating interpretive uncertainty.
- Critics including Washington Legal Foundation warn the rule's broad scope may admit unreliable AI evidence.
The Federal Advisory Committee on Evidence Rules has introduced Federal Rule of Evidence 707 to set clear standards for admitting evidence generated by artificial intelligence and other machines.
Unlike traditional expert witness testimony under Rule 702, Rule 707 proposes admitting machine-generated evidence on reliability alone. It applies Rule 702’s reliability criteria: helpfulness to the factfinder, sufficient data, and reliable methodology and application. However, it excludes evidence from "simple scientific instruments," a term left undefined, which could lead to uncertainty about what evidence is covered.
The public comment period ran from June 10, 2025, through February 16, 2026. The committee voted on the proposal in May 2026. If approved by the Supreme Court and Congress, the rule is expected to take effect on December 1, 2027.
Critics have raised concerns. The Washington Legal Foundation (WLF) criticized the broad definition of “machine-generated” evidence. Zac Morgan, senior counsel at WLF, warned, “Rule 707 is not ready for primetime,” highlighting risks of admitting unreliable AI outputs beyond sophisticated systems.
The American Association for Justice urged narrowing the scope to explicitly cover AI and machine learning-generated evidence and exempt routine technologies to avoid unnecessary litigation.
Judges will face challenges in evaluating AI evidence. Judge Paul W. Grimm has emphasized the need to scrutinize such evidence for bias and reliability issues. This complexity arises from AI’s opaque operations and potential for errors.
Legal experts note the rule’s phrasing — "the court may admit" — offers discretion but could be interpreted as lowering admissibility standards compared to the rigorous expert testimony traditionally required. Courts will need to balance flexibility with rigorous evaluation to maintain evidence integrity in the digital age.
By the numbers:
- February 16, 2026 — public comment period ended
- May 2026 — Advisory Committee vote on Rule 707
- December 1, 2027 — potential effective date if adopted
Yes, but: While Rule 707 aims for clarity, the undefined exclusion of 'simple scientific instruments' and discretionary language may create ambiguity and inconsistent application in courts.
What's next: The Supreme Court and Congress must review and approve Rule 707 before the expected December 2027 implementation date.