AI Handles the Thinking. What Handles the Proof?

On March 23, 2026, Above the Law published an article arguing that AI won't replace lawyers but could create a critical shortage of good ones. The piece captures the conversation happening across the legal industry right now: what does AI mean for the profession?

It's the wrong question. Or at least, it's not the most urgent one.

The more pressing issue isn't whether AI replaces attorneys. It's what happens to evidence integrity when AI accelerates every other part of litigation. Review is getting faster. Sharing infrastructure has not kept up. And the gap between the two is where risk accumulates.

The review problem is solved. The sharing problem isn't.

AI tools are already embedded in legal workflows. Claude, GPT, Copilot, and purpose-built legal AI platforms are reviewing documents, summarizing depositions, and drafting motions. Every major firm is either experimenting with or actively deploying AI-assisted review. The ABA's 2024 Legal Technology Survey found that 30% of attorneys now use AI tools in their practice, nearly triple the figure from the year before.

The technology works. AI can identify the key documents in a production, flag privilege issues, extract relevant facts, and organize materials by issue in a fraction of the time manual review requires.

But once the AI identifies the key evidence, what happens next?

That evidence gets emailed as an attachment. It gets shared via Dropbox or Google Drive. It gets uploaded to a generic cloud storage service with a link that anyone with the URL can access.

No integrity verification. No viewer attribution. No chain of custody documentation. No court-ready authentication.

The faster AI makes review, the more evidence gets shared across parties, and the bigger the authentication gap becomes. AI didn't create this problem. But it is accelerating the volume of evidence moving through workflows that were never built to handle authentication at scale.

FRE 902(13) was written for this moment

Federal Rule of Evidence 902(13), effective since December 1, 2017, allows electronic records to self-authenticate when accompanied by a written certification from a qualified person. The rule was designed for exactly this scenario: proving that digital evidence hasn't been altered between collection and presentation.

The mechanism is straightforward. A cryptographic hash function — specifically SHA-256 — generates a unique fingerprint of the file at the time of upload or collection. That fingerprint is recorded. At any later point, the same hash function can be run again. If the output matches, the file is identical to the original. If it doesn't, the file has been altered.

A qualified person then certifies under penalty of perjury, pursuant to 28 U.S.C. § 1746, that the system's process produces an accurate result. The certification is filed with the court at least 14 days before trial. Opposing counsel can challenge authenticity, but the proponent doesn't need a live witness to lay foundation.

This isn't new law. It's existing law that most firms aren't using because they don't have the tooling to generate the hash, maintain the chain of custody, and produce the certification in a format courts expect.

The Advisory Committee Notes to Rule 902(13) specifically state that the rule was intended to reduce the burden of authentication for electronic records where the integrity of the system can be demonstrated through certification rather than testimony. The infrastructure to support that certification is the missing piece.

Viewer attribution is the missing layer

AI can review a document. It can extract every relevant fact, flag every privilege issue, and organize an entire production by topic. But when evidence is shared with opposing counsel, experts, or clients, there is typically no record of who actually viewed it.

Viewer identity watermarking addresses this. When visual evidence is viewed through Attested, the viewer's verified email address is displayed as an overlay on every frame. The watermark is visible in any screenshot or screen recording taken during the viewing session. Removing it from a captured image would require deliberate editing, which itself is evidence of intent to circumvent security controls.

The watermark is a deterrent, not a guarantee. It is a CSS overlay rendered in the browser, not a forensic mark embedded in the file data. For document formats where text can be copied regardless of visual overlays, Attested relies on the access audit trail rather than watermarking.

The access audit trail provides a parallel record. Every view, download, and share is logged with the viewer's verified identity, timestamp, IP address, and device information. Together, the watermark and the access log answer the question that email and cloud storage cannot: who viewed this evidence, and when.

AI review + authentication in one pipeline

The ideal workflow for an attorney handling digital evidence in 2026 is not complicated. It is a single pipeline with four steps.

Upload the evidence. The system hashes the file with SHA-256 at ingestion, creating an immutable fingerprint. Get an AI-generated analysis — summary, key facts, entity extraction, timeline, and privilege flags — that saves hours of manual review. Share the evidence securely with viewer attribution, so every recipient's identity is embedded in the content and every access event is logged. Generate a court-ready FRE 902(13) certificate that includes the hash, the access log, and the AI analysis as part of the system's processing record.

This is what Attested does. Upload, review, share, certify. Every step logged. Every viewer identified. The AI review results are timestamped and included in the FRE 902(13) certificate as part of the documented processing record, which supports authentication of both the evidence and the analysis performed on it.

This is not about replacing lawyers. Attorneys still make the judgment calls: what to share, with whom, under what protective order, and with what conditions. AI handles the analytical work. The authentication infrastructure handles the proof that the evidence is intact and that every viewer is accounted for.

AI makes lawyers faster. Evidence authentication makes them defensible. The firms that figure this out first will have a structural advantage in every case they touch.