Question · FRE 707 (proposed)
How do I lay a reliability foundation for an AI model's output?
Proposed Rule 707 would make AI or machine-generated evidence offered without an expert meet the Rule 702 reliability requirements.
01 / The answer
The short answer
Show that the system is reliable and was used reliably in your case. In practice, document what the model is and which version ran, the data it was built and validated on, its known error rates or test results, the inputs behind the specific output, and that the run can be reproduced. Where you can, have a qualified expert explain these points; the Advisory Committee's revised Rule 707 draft would presume one is needed. Give the other side early notice and access so it can test the result.
02 / The rule
What FRE 707 (proposed) requires
Proposed Rule 707 would make AI or machine-generated evidence offered without an expert meet the Rule 702 reliability requirements. It was published for comment in August 2025, drew mixed comments, and in May 2026 the Advisory Committee chose to revise and keep studying it rather than send it forward, so it has no effective date.
- 01As published: machine-generated evidence offered without an expert must satisfy Rule 702(a)-(d) if it would need Rule 702 when given by a witness.
- 02Revised draft: proponent shows the AI evidence helps the trier of fact, rests on sufficient data, uses reliable methods, reliably applied.
- 03Revised draft: ordinarily provide an expert to explain how the AI system reliably produced the evidence; other proof only in exceptional circumstances.
- 04Revised draft: give all parties reasonable notice of intent to offer AI-produced evidence.
03 / Records
Records to keep
- 01Validation and error-rate records for the AI system
- 02Description of training and input data sufficiency
- 03Reproducible run records showing reliable application to case facts
- 04Pretrial notice served on opposing parties
Optional evidence · where Planisphere fits
Keep a record anyone can check
When your team meets a duty, Planisphere can seal the event it reports as a receipt.
A disclosure shown, a review signed, a model change approved: the event your system reports.
The event becomes a sealed receipt, dated and attributed to whoever is signed in.
Paste the receipt into Verify. No account or key.
A record supports your compliance evidence process. It is not a legal finding or a certification. Systems that call Planisphere's gate API can also hold or refuse an action before it runs.
SOURCES
Go back to the source
Administrative Office of the U.S. Courts
Administrative Office of the U.S. Courts
RELATED
Related: FRE 707 (proposed): the full guide · Is AI-generated evidence admissible in federal court? · What is proposed Federal Rule of Evidence 707? · When does FRE 707 take effect? · Can an audit record prove my AI tool is reliable enough for court?