Artificial Intelligence not only requires practitioners to adapt their methodology and strategy—it also demands that the rules governing our practice shift as well. 

Changing core procedural rules, like the Federal Rules of Evidence, is never easy, but staying informed and ready to adapt is essential.  Proposed FRE 707, “Machine-Generated Evidence,” captures this tension: it has proven particularly divisive and remains unfinished.  

The Proposed Rule

Proposed Federal Rule of Evidence 707 is one of the most closely watched rulemaking developments.  The proposed rule would require machine-generated evidence offered without an expert witness to satisfy Rule 702(a)-(d) if it would otherwise be subject to Rule 702 as expert testimony if testified to by a witness.  The rule does not apply to the output of simple scientific instruments such as thermometers and scales.

Practically, Proposed Rule 707 would ensure that AI or other machine-generated output offered without a sponsoring human expert would have to meet the same Daubert-derived reliability criteria highlighted in Rule 702: (1) it must assist the trier of fact, (2) it must be based on sufficient facts or data, (3) it must be the product of reliable principles and methods, and (4) it must reflect application of those principles and methods to the facts of the case. 

This proposal intends to close a loophole: a party would not be able to avoid Rule 702’s reliability gatekeeping by offering machine output directly into evidence instead of through a human expert’s opinion.

The Advisory Committee Report’s example of the type of applicable evidence was an AI diagnostic report in a medical malpractice case concluding a physician missed a diagnosis.  However, this proposed rule’s application would extend to all litigation involving algorithmic outputs, forensic analysis, or AI-assisted discovery tools.

Current Status

Proposed Rule 707 was published for comment and then sent back for revision but has not yet been adopted.

Divided Opinions on Rule 707

While on paper Proposed Rule 707 seems to simply enforce the already widely established Daubert standard to machine-generated evidence, there is widespread concern about the rule’s scope and its downstream effects.

The American Association for Justice’s formal comment argued that the rule was overly broad and could subject routinely admitted evidence like geolocation data and surveillance footage to heightened standards.  This slippery slope-style argument highlights the risk that Proposed Rule 707 would make it harder to admit long-accepted forms of evidence rather than function as a gatekeeper preventing novel AI outputs from circumventing the rules.

On the other hand, some commentators question whether the rule is too narrow given that Proposed Rule 707 would apply only to evidence the proponent acknowledges was AI-generated.  As a result, this does not help courts address disputed authenticity—or actual deepfakes—where a party denies AI involvement at the outset.

Another concern was that the rule would require expensive technical experts to analyze complex AI systems, resulting in an “uneven playing field between under-resourced litigants and those with unconstrained resources.”  Others contend it could also be used as a delay tactic instead of a quality-control mechanism.

As an alternative to creating a new standalone rule, the NYC bar recommended folding Rule 707 into existing discovery mechanisms such as Civil Rule 26 or Criminal Rule 16.  The DOJ believed that AI-generated evidence was already covered by Rule 702 and that Proposed Rule 707 addressed speculative future needs rather than a present problem. 

Takeaways

Whether this particular rule comes to fruition or not, Proposed Rule 707 symbolizes the importance of ensuring the federal rules still serve their purpose in the age of machine-generated evidence.  It also serves as a reminder that litigators must be prepared with validation and explainability materials for AI-based evidence given that Rule 702’s underlying reliability principles will likely inform any final rule.  Practitioners should look out for the next Advisory Committee meeting for a revised draft of the rule and to learn of its ultimate fate.