Why Brady Violations Still Happen in 2026

Why do Brady violations persist more than sixty years after Brady v. Maryland? The answer is uncomfortable: they happen because nothing stops them. A 2010 study of 35 current and former state prosecutors across 7 offices found that most violations stem from negligence, cognitive bias, or simple mistakes rather than intentional misconduct, and that cognitive bias "cannot be overcome by force of will, good intentions, or even training," according to NACDL's analysis of the study. Office culture and policy, not the law itself, drive pretrial disclosure practices.

The materiality trap sits at the center of it. Prosecutors reflexively apply the appellate standard (reasonable probability of a different outcome) to define their pretrial obligations, even where that isn't the law in their jurisdiction, and many offices have no clear policy at all. The Supreme Court has said pretrial disclosure duties are broader than the post-trial reversal standard, and prosecutors are poor judges of what counts as material to the defense, as The Appeal's explainer on Brady violations lays out. One line prosecutor captured the timing dynamic bluntly: "If he did, why should I help the defense win its case by turning over information to discredit my witness before I have to?"

Then there's the digital-evidence mismatch. Brady was designed for a 1963 world of a handful of paper documents, but police now pass data to prosecutors "usually terabytes at a time," and prosecutors often respond by dumping entire case files, knowing full well that defense counsel will fare no better at finding exculpatory items in bulk data. NYU Law Review argues this puts the adversarial system on shaky ground.

Accountability is close to nonexistent:

• A Northern California Innocence Project study of roughly 600 misconduct cases found the state bar publicly disciplined only 1% of prosecutors.
• Only one state requires prosecutors to maintain Brady lists, and disclosure practices "often depend on the changing whims of elected officials."
• Thousands of officers with misconduct histories continue to work and testify while many prosecutors keep no Brady list at all.
• Researchers rarely found retrospective reviews, spot-checks, or random audits of disclosure failures in prosecutors' offices.

Two framings compete for why this continues. NACDL's study treats most violations as unintentional, driven by bias and negligence. The Appeal points to a win-at-all-costs mentality and quotes former Ninth Circuit Judge Alex Kozinski saying prosecutors "consider [Brady violations] feathers in their caps." Both are probably true, and the practical consequence is the same either way: you cannot assume the prosecution will screen this for you. Teams that build their own Brady list research workflow and run grounded, citation-backed review of the discovery dump find the favorable evidence before anyone asks.

The Digital Evidence Dump Broke the Disclosure Model

Brady v. Maryland was decided in 1963, when discovery meant a handful of paper documents. Sixty years later, police hand prosecutors data "usually terabytes at a time," and the doctrine simply wasn't built for it.

The result is a disclosure model that fails in both directions. According to NYU Law Review, prosecutors increasingly respond to massive digital dumps by turning over entire case files, "knowing full well that defense counsel will fare no better" at finding exculpatory material buried in bulk data. That practice, the authors argue, puts the adversarial system on shaky ground. A dump is not a disclosure.

Then there's the Brady list, the mechanism meant to track officer credibility. Minnesota Law Review research finds that only one state requires prosecutors to maintain Brady lists at all, while thousands of officers with misconduct histories keep working and testifying. Disclosure practices vary enormously and often depend on the changing whims of elected officials.

For defense teams, that combination changes the job:

• Assume no Brady list disclosure is coming, and build a Brady list research workflow of your own.
• Treat the bulk dump as a review problem you have to solve, not a file transfer you received.
• Work from diarized transcripts, OCR'd documents, and inconsistency signals (report vs. bodycam, statement vs. call audio) so someone actually mines the terabytes.
• Hold every tool to an auditability standard: traceable citations, visible sourcing, human review before anything becomes an argument.

There's a newer wrinkle too. Thomson Reuters describes a "parity problem" in which defense counsel use AI review tools while prosecutors may not, producing what it calls an uneven courtroom. Note the source: that's vendor-sponsored content for CoCounsel Legal, and its efficiency numbers are self-reported. The framing still cuts both ways. Prosecutors' offices already run 80 to 150 active cases per attorney amid staffing shortages, per the same vendor study. A defense team with grounded, citation-backed review can find overlooked favorable evidence before the prosecution does.

That's the practical posture for 2026. Don't wait on the other side's diligence. Build your own record, and make it one you can point to.

Build Your Own Brady Review: The Five Categories Checklist

Here's the uncomfortable arithmetic: prosecutors screen your Brady material under a materiality standard the Supreme Court has said is wrong for pretrial disclosure, and they do it while carrying 80 to 150 active cases per attorney. If you wait for their screen to catch what matters to your client, you're gambling with someone else's incentives.

A 2010 study of 35 prosecutors across seven offices found most Brady violations stem from negligence and cognitive bias rather than intent, and that bias "cannot be overcome by force of will, good intentions, or even training." Office culture, not the law, drives what gets disclosed. That's why the defense-side answer is simple: never rely on prosecutor screening. Run your own review against the five categories of material most often withheld, as catalogued by The Appeal:

• Alternative perpetrators. Anything pointing to a third party: other suspects in police reports, 911 caller descriptions that don't match your client, tips logged and never run down.
• Giglio material. Prior inconsistent statements from prosecution witnesses, especially ones that surface only when you compare the grand jury testimony against the bodycam or the interrogation footage.
• Witness deals. Any consideration for testimony: leniency letters, immunity agreements, dropped charges for a codefendant, favorable treatment on pending cases.
• Police credibility. Evidence planting, falsified reports, prior misconduct by the testifying officers. Only one state mandates prosecutor-maintained Brady lists, and disclosure practices vary with "the changing whims of elected officials," per Minnesota Law Review research, so you can't assume an officer's history will arrive in discovery. A structured Brady list research workflow for identifying officer-witnesses with misconduct histories is now part of the job.
• Lab issues. Protocol changes, analyst misconduct, accreditation problems. In one case The Appeal documents, a defense team learned of a DNA protocol change only by accident; the conviction was reversed in October 2017 because the change had invalidated the analyst's testimony.

Why look beyond the case file? Because under Kyles v. Whitley, the disclosure duty reaches material held by police, forensic investigators, and other experts, not just what the line prosecutor personally knows. The prosecutor handling your case may genuinely not know what the detective or the lab knows. That gap is yours to close.

The practical problem is volume. Police now pass data to prosecutors "terabytes at a time," and prosecutors increasingly respond by dumping entire case files, knowing full well that defense counsel will fare no better at finding the exculpatory needle in the bulk transfer. (The NYU Law Review piece is at nyulawreview.org if the link above misroutes; the argument is that bulk dumps put the adversarial system "on shaky ground.")

This is where a grounded review workspace earns its keep. In Evidence Copilot, you upload the discovery dump (bodycam, interrogation video, 911 and jail calls, scanned reports), get diarized transcripts and OCR'd text, and the auto-signals flag candidate inconsistencies like report-versus-bodycam conflicts before you think to ask. Every answer cites the exact video second or PDF page, and everything stays marked for human review, because the tool surfaces the material; you make the Brady call.

Work the five categories on every case. The record you preserve early is the one that wins the Kyles argument later.

Your Defense-Side Workflow: Brady Lists, Bulk Discovery, and Auditability

You can't count on the prosecution to hand you the Brady material. Only one state requires prosecutors to maintain Brady lists, and disclosure practices often depend on "the changing whims of elected officials," according to research in the Minnesota Law Review. So the review is yours to run.

Start with independent officer research. Before you depose or cross a testifying officer, run a structured Brady list research workflow that maps the officer-witness to known misconduct histories. Build it into intake, not the week before the preliminary hearing.

Then mine the dump itself. Prosecutors increasingly turn over entire case files "knowing full well that defense counsel will fare no better" at finding exculpatory items in terabytes of data, per the NYU Law Review. A technology-assisted review workflow changes that math:

• Timestamped, diarized transcripts of bodycam, interrogation, and jail-call audio so you can search spoken words, not just skim footage.
• OCR across scanned reports and photo exhibits so text becomes searchable and citable.
• Report-vs-bodycam and statement-vs-call inconsistency signals that surface candidate impeachment material before anyone asks.
• Cross-exam drafts built from a flagged signal, with both sides' quotes attached.

Demand auditability from any AI tool you adopt. Ask whether it produces a complete audit trail, whether every output traces to an authoritative source, and how it handles hallucinations, the framework Thomson Reuters outlines for evaluating legal AI. Grounded, citation-backed outputs marked for human review are the floor, not the ceiling. Every AI answer in Evidence Copilot's workspace cites the exact second of video or page of the PDF it came from, and nothing is auto-applied.

The parity problem cuts your way. Prosecutors manage 80 to 150 active cases per attorney amid staffing shortages, per vendor-reported Thomson Reuters data, and in narrow-compliance jurisdictions line prosecutors control disclosure timing. A defense team that reviews faster and grounds every citation can find overlooked favorable evidence, and preserve the record, before the state does.

Book a 20-minute walkthrough at evidencecopilot.com/book, or start on the Solo plan at evidencecopilot.com/signup.

Exploit the Parity Problem Before the Prosecution Does

Prosecutors' offices run 80 to 150 active cases per attorney, and in narrow-compliance jurisdictions the timing of disclosure is left to the line prosecutor's discretion. That combination is not a reason to wait. It is a reason to move first.

The disclosure model itself is under strain. Police now hand data to prosecutors "usually terabytes at a time," and offices often respond by turning over the entire case file, knowing full well that defense counsel will fare no better at finding the exculpatory needle. Bulk transfer is not disclosure. It is a review problem handed to you unsolved, and it puts the adversarial system on shaky ground.

You can solve it faster than the other side. A grounded review workflow that produces diarized transcripts, OCR'd document text, and automatic inconsistency signals (report versus bodycam, statement versus call audio) lets your team actually mine the dump instead of skimming it. Build the review around the five categories of frequently withheld material:

• Evidence suggesting an alternative perpetrator
• Giglio material: a witness's prior inconsistent statements
• A witness's motive to lie, such as a deal for leniency
• Information casting doubt on police credibility
• Information casting doubt on crime lab technicians, including protocol changes

That last category is not hypothetical. A conviction was reversed in October 2017 after the defense accidentally learned of a DNA protocol change that invalidated the analyst's testimony. Nobody disclosed it. The defense found it by luck.

Do not count on a Brady list disclosure arriving either. Only one state requires prosecutors to maintain Brady lists, and disclosure practices often depend on the changing whims of elected officials. Run your own officer research through a structured Brady list research workflow, and pair it with in-case party records so officer-witness histories surface against the file in front of you.

Speed matters for one more reason: preservation. When you identify favorable evidence early, you can document it, cite it to the exact source moment, and build the record your motion practice depends on. Evidence Copilot is a grounded review workspace built for that sequence, where every AI answer carries a citation into the exact second of video or page of the PDF and every output is marked for human review. If you want to see the workflow on your own discovery, book a 20-minute walkthrough at evidencecopilot.com/book.