Queen opened its most famous six-minute epic by asking, in effect, whether what we are looking at is real life or only a fantasy — whether we are awake, or caught in something we cannot climb out of. For almost the entire history of litigation, that was a question we never had to ask about the evidence in a case. A photograph was a photograph. An email was an email. A recording was a recording. The authenticity fights were about chain of custody and foundation, not about whether the thing in front of us had ever existed at all.
That era is over, and the rules of evidence are scrambling to catch up. This post is about what happens to discovery, and to your evidence, when generative AI can manufacture a document, a photo, a voice, or a video that survives a first, second, and third look.
Is there really no escape from reality?
Freddie Mercury's narrator insisted there was no escape from reality even as the song dissolved into pure invention. Litigators are now living the inverse problem: the invention is getting good enough that reality no longer announces itself. A convincing email thread can be generated in seconds. A photograph can be altered so that the manipulation leaves no seam a human eye will catch. A voice can be cloned from a few seconds of audio. The comfortable assumption that a document is, at a minimum, a real document — that it was written by someone, at some time, and merely needs its authorship and relevance established — is an assumption that no longer holds.
This matters long before trial. It matters the moment a client hands you a text message or email that supposedly proves the other side's bad faith. It matters when the other side produces one that supposedly sinks your case. The question that used to be reserved for the rare forgery is becoming a routine part of evaluating any piece of electronic evidence: is this thing what it claims to be, and how would I prove it either way?
Open Your Eyes
The rulemakers have seen this coming. In August 2025, the Committee on Rules of Practice and Procedure released a proposed new Federal Rule of Evidence 707 for public comment. The rule would take machine-generated evidence offered without a supporting expert and hold it to the same reliability standard that governs expert testimony under Rule 702 — meaning the output would have to rest on sufficient data, reliable methods, and a reliable application of those methods to the facts before a court could let a jury see it. (The original proposed language began: “When machine-generated evidence is offered without an expert witness and would be subject to Rule 702 if testified to by a witness, the court may admit the evidence only if it satisfies the requirements of Rule 702(a)–(d). This rule does not apply to the output of simple scientific instruments.”) The comment period closed on February 16, 2026. When the Advisory Committee on Evidence Rules took the proposal up on May 7, 2026, however, it declined to advance Rule 707 as written — revising the rule and holding it for further study rather than sending it up the chain.
So the machine-generated-evidence rule has stalled, and even if a version eventually clears every remaining hurdle — the Standing Committee, the Judicial Conference, the Supreme Court, and Congress — the earliest it could take effect is December 1, 2027. Rule 707 is a real step, but read the fine print on its limits. It reaches evidence the proponent admits was generated by a machine. It does very little about the harder case — the photograph or recording someone offers as genuine while the other side insists it is a fake. For that problem, the rule-makers are looking at a different tool.
Will you let me go?
Notably, at that same May 2026 meeting the committee focused its energy on the substance of a proposed Rule 901(c), aimed squarely at deepfakes and creating a burden-shifting process. Under the version the Advisory Committee has been weighing, a party challenging electronic evidence as AI-fabricated cannot simply say the word "deepfake" and force the other side to prove a negative. The challenger must first come forward with evidence sufficient to support a finding that the item was in fact manipulated or fabricated. Only then does the burden shift to the proponent, who must establish that the evidence is more likely than not authentic — a meaningfully higher bar than the traditional "enough to support a finding" standard.
A Federal Judicial Center survey released in March 2026 confirmed the direction of travel: a bare assertion that something is a deepfake would not be enough to trigger the heightened process. The system is trying to thread a needle — take fabrication seriously without handing every litigant a free, evidence-free objection to anything inconvenient. Whether it succeeds is something we will, fittingly, only find out over time.
Any way the wind blows
Here is the part that matters most for businesses and the lawyers who advise them: none of this is in force yet, and the earliest any of it could be is late 2027. The disputes are arriving now. That gap is not a reason to wait; it is a reason to build good habits ahead of the rules.
Practically, that means treating provenance as something you protect rather than assume. Preserve native files with their metadata rather than relying on screenshots and printouts, because the metadata is often what lets you prove — or disprove — that a document is genuine. Widen litigation holds to capture the surrounding digital record: the account it came from, the transmission data, the device. When you receive a damaging document in discovery, ask early whether you can authenticate it independently, and when you rely on your own evidence, ask whether you could carry a more-likely-than-not burden if the other side cried fabrication. The firms that handle this well will be the ones that stopped assuming a document is real simply because it looks real.
Gotta leave you all behind and face the truth
The song wandered through fantasy and then landed, quietly, on the idea that reality was still there waiting. Litigation is heading the same way. The tools for manufacturing evidence are improving faster than the rules meant to govern them, and for the next few years the burden of separating the real from the fabricated will fall on lawyers who know to look, know what to preserve, and know how to make — and meet — an authenticity challenge. The rules will catch up. Until they do, the eyes doing the looking are yours.
This post is part of an ongoing series on how AI is reshaping the practical realities of litigation and law-firm practice. In earlier posts we looked at AI-generated fabrications from the deepfake angle and at the sanctions courts are imposing when AI invents authority; here we turn to what all of that means for the evidence in your own case. And your author apologizes for not being able to find a way to use “Scaramouche,” “Galileo,” “Figaro,” or “Magnifico” in this article.
A Word About Silver Cain
Silver Cain PLC represents businesses in complex commercial and real estate litigation in Arizona and beyond. When Rebecca Cain and I founded the firm, we built it around direct partner involvement, senior trial-level judgment, and a disciplined, clear-eyed approach to the evidence a case actually turns on — including the growing set of questions about what that evidence really is. If the issues in this post are relevant to your business, or to the firms you retain, we are glad to have that conversation.
Leon Silver is an AV-rated trial lawyer at Silver Cain PLC, focused on commercial and real property disputes since 1989. Reach him at lsilver@silvercain.com.
