Deepfakes, AI, and the Future of Text Message Evidence
Two recent developments — a New York court ruling and a proposed federal evidence rule — both point the same direction: courts are getting more skeptical of digital evidence that can't be independently verified. That has real implications for how text message evidence should be handled, even though neither development is specifically about text messages.
Matter of M.S. (M.H.) — video authentication tightens
In a majority opinion dated February 17, 2026, authored by Chief Judge Rowan Wilson, the New York Court of Appeals held that video clips central to a child-protective proceeding had not been properly authenticated. The person who allegedly created the videos didn't testify, and much of the authentication testimony that was offered was hearsay. The majority specifically noted that “the increasing prevalence of ‘deepfake’ videos has only rendered the method of matching circumstantial details in a video to personal observations a more suspect form of authentication.” Strong dissents in the case warned this standard creates a very high bar going forward, one that could make it harder to hold abusers accountable using video evidence. This case is specifically about video, not text messages — but it signals a broader judicial mood: circumstantial, “this looks right to me” authentication is losing ground to more rigorous standards.
Proposed FRE 707 — machine-generated evidence
Separately, a proposed new federal rule — Rule 707 — would have required machine-generated evidence offered without an accompanying expert witness to meet the same reliability standards as expert testimony under Rule 702(a)–(d). Public comment on the proposal closed in February 2026. In its report dated May 17, 2026, the Advisory Committee on Evidence Rules recommended no action on the rule “at this time,” and the Standing Committee agreed at its June 2026 meeting — sending Rule 707 back for revision and further study alongside a related, separate proposal specifically about deepfakes. As of this writing, Rule 707 does not exist and has no effective date. The underlying concern that motivated it — how courts should treat machine-generated or algorithmically- produced evidence — hasn't gone away just because this particular draft was shelved; it's likely to resurface in a revised form.
What this means for text message evidence specifically
A text message export isn't a deepfake video or an AI-generated document — it's a record of something that was actually typed and sent. But the direction both of these developments point is the same: the more a piece of digital evidence relies on someone simply asserting “this is real” without a verifiable trail behind it, the more scrutiny it's likely to face. A native export with intact timestamps and metadata is in a much stronger position than a screenshot with neither, for exactly this reason.
Verity is a forensic and behavioral analysis tool, not a law firm and not a certifying authority. Nothing here is a claim that Verity's output satisfies any specific evidentiary rule, survives a particular legal challenge, or constitutes expert testimony. Legal rules and case outcomes described here can change, and this is general information, not legal advice — talk to an attorney about how any of this applies to your specific situation.
Work from a native export, not a screenshot
Upload your export and get a free health score in minutes — the analysis runs in your browser, nothing is uploaded.
Get your free health score →