Netanyahu's Café Video Shows Why "I Saw It on Video" No Longer Counts as Evidence | Podcast

An A.I. chatbot looked at a real video of Benjamin Netanyahu sitting in a café — verified footage, confirmed location, actual event — and declared it one hundred percent deepfake. Not fifty-fifty. Not suspicious. One hundred percent fake. The video was real. The A.I. was wrong.

That matters to you because every piece of video

That matters to you because every piece of video evidence you encounter now lives in this same gray zone. According to Reuters, the footage showed Netanyahu in a coffee shop, and journalists confirmed the location using file imagery from the scene. But Grok, an A.I. chatbot, flagged it as fabricated — citing static coffee levels in the cup and unnatural lip movements. A sitting prime minister had to post what amounted to a proof-of-life video. If authentic footage of a world leader can't survive an A.I. detection scan, what happens when a video clip becomes the key exhibit in your next case?

Start with the courtroom. In November of last year, the Advisory Committee on Evidence Rules met to consider a proposed Rule 901(c). That rule would govern how courts handle electronic evidence that might've been fabricated or altered by A.I. Then this past August, the Judicial Conference released a separate rule — Rule 707 — for public comment. But critics spotted a gap wide enough to drive through. Rule 707 only applies when the person introducing the evidence admits it was A.I.-generated. It does nothing when authenticity itself is the fight. And authenticity is almost always the fight now.

That gap feeds something researchers call the liar's dividend. A bad actor points at a legitimate recording and says — that's a deepfake. Suddenly the jury isn't weighing the merits of the case. They're stuck debating whether the evidence is even real. The trial becomes a trial about the tape before it becomes a trial about the truth.

So who pays for that fight? Under the Daubert standard, judges act as gatekeepers. They evaluate whether an expert's methods are testable, peer-reviewed, and generally accepted. Most proprietary deepfake detectors offer no audit trail. They can't clear that bar. That means both sides hire competing forensic experts, and the cost spirals. A well-funded defendant can demand hearing after hearing. A solo investigator or a small plaintiff's firm can't keep up.