Every face on the screen was synthetic. The lesson for litigators is uncomfortable. If a global company’s own staff can’t tell a live executive from a generated one on a Teams call, a jury looking at a grainy exhibit stands no chance.
That’s the discovery pile now. Photos, voice memos, security-cam clips, PDFs, chat logs — any of it could be generated, edited, or spliced by a model, and the person offering it into evidence may have no idea.
The Problem Sitting on Every Litigator’s Desk
Discovery used to run on a simple assumption: a document, a photograph, or a recording was either real or it wasn’t, and a witness who was there could usually tell you which. That assumption has stopped holding. Synthetic media is cheap, fast, and good enough to fool the person it depicts.
Volume matters as much as quality. Fabricated exhibits used to be rare enough that a suspicious one stood out on its own. Now anything digital arrives under a small cloud of doubt, and the doubt lands on the lawyer who wants to use it, or the lawyer trying to keep it out.
A few things show up over and over in intake and production:
- Doctored photographic evidence. A crash-scene photo where the position of a vehicle has been nudged, a bruise deepened, or a timestamp swapped. The image still passes a casual eye test.
- Fabricated audio and video. A voicemail the client swears they never left. A doorbell clip showing a person who was, by every other record, three states away.
- Synthetic documents. Invoices, medical records, text-message screenshots, even signed letters produced by a model in under a minute, formatted to match the real thing.
- AI-assisted expert output. Charts, reconstructions, and analyzes generated by tools whose training data and error rates nobody in the courtroom can inspect.
What Firms Are Actually Doing About It
The firms getting ahead of this are handling synthetic-media risk as a workflow problem, not a rule-of-evidence puzzle to be solved on the eve of trial. Judges and rules committees are moving in the same direction. The Judicial Conference recently approved a new Federal Rule of Evidence aimed squarely at machine-generated output, and trial courts are already being asked to draw lines around deepfake exhibits. A few habits are showing up consistently:
- Preserve the source, not the screenshot. A screenshot of a text is almost useless now. Firms are collecting the device, the cloud export, the carrier record — anything with metadata a forensic examiner can work with.
- Chain of custody, from intake forward. Client-supplied media gets logged the day it arrives: who sent it, in what format, with what hash. Waiting until the deposition to ask where a video came from is waiting too long.
- A named expert on speed dial. Digital forensics used to be a trial expense. It’s becoming an intake expense. A short pre-litigation review can flag a doctored exhibit before a demand letter goes out on it.
- Deposition questions rebuilt for provenance. Not just “is this your text?” but what device, what app, was it forwarded, was it edited, do you still have the original. The old script assumed most of that was obvious.
- Written policies for the firm’s own AI use. Firms using generative tools for demonstratives or summaries are documenting which tool, which prompt, and which human reviewed the output, because opposing counsel will ask.
None of this is unusual. It’s the discipline litigators already apply to a paper file, moved earlier and applied to formats that didn’t exist a decade ago. Plaintiffs’ firms that handle a steady volume of injury and accident work — the kind of practice a personal injury firm sees every week — are especially exposed here, because so much of the proof (dash-cam clips, phone photos, doorbell video, texted admissions) now arrives in exactly the formats a model can generate.
The practical takeaway isn’t memorizing a rule that may change next cycle. It’s building the intake and preservation habits now, so that when new guidance lands — federal, state, or bar-association — the firm’s files already meet it.
The exhibits that survive the next few years of motion practice will be the ones whose provenance was documented before anyone knew there was going to be a fight about it.
Founder & Editor at Durofy
