Why Brady Violations Keep Happening in 2026

If you assume the prosecution's disclosure process will surface exculpatory evidence, the numbers say otherwise. A study of 386 Brady violations between 2004 and 2022 found that 49% occurred in homicide cases, and 66% of prosecutorial violations were committed in bad faith. The Death Penalty Information Center has identified more than 200 death-sentenced individuals whose convictions were overturned for prosecutorial misconduct involving a Brady violation.

The uncomfortable truth is that most violations grow out of structural failure, not a single villain. Professor Adam Gershowitz describes the system as "a jigsaw puzzle with a thousand tiny pieces" in which no one is really in charge, and Professor Jennifer McAward calls the prosecutorial team "just a collection of parts that do not work well together." When nobody owns the disclosure duty, evidence falls through the cracks.

The caseload math explains a lot. Fort Worth misdemeanor prosecutors carry between 1,200 and 1,500 matters apiece, and large DA offices report vacancy rates of 15 to 33%. A prosecutor buried in that many files cannot meaningfully review what police send, let alone chase what they didn't.

The most important number for your practice is this one: in 51% of successful Brady claims, the withheld material never reached the line prosecutor at all. The failure happened upstream, between the investigating officer and the DA's office. That means waiting on the prosecution's disclosure is waiting on a pipeline with a known 50% leak rate. The materiality standard requires only a "reasonable probability" that the evidence would matter, but you can't argue materiality for a document you never saw.

The practical takeaways for defense teams:

• Assume the discovery dump is incomplete. Cross-reference police reports, bodycam, jail calls, and 911 audio yourself rather than trusting the state's file.
• Watch for impeachment material. In 45% of violation cases, the suppressed evidence could have impeached a critical prosecution witness.
• Keep the duty alive. Because disclosure is ongoing, a Brady motion can be filed after openings, mid-cross, or before sentencing.
• Build a repeatable research process. A structured Brady list research workflow turns officer-credibility digging into a checklist instead of a scramble.

This is also why the 2026 NACDL report on defender AI adoption matters here. It frames competent, secure tooling as part of the parity problem: prosecutors are adopting AI, and defenders who review discovery by hand are reviewing less of it. Evidence Copilot's grounded evidence-review workspace exists for exactly this gap, comparing every uploaded file against the rest of the case and flagging report-versus-bodycam inconsistencies before anyone thinks to ask.

The evidence that clears your client is often already in the discovery dump. The question is whether anyone on your side reads all of it before the state does.

Brady vs. Giglio: What to Ask For and When

Most Brady motions fail because the defense asks for the wrong bucket. The officer's disciplinary file isn't a Brady claim, and treating it like one gets your motion denied before the judge even reaches the merits.

A Brady violation concerns exculpatory or impeachment evidence in the prosecution's possession: evidence that supports innocence, undercuts the state's theory, or impeaches a key witness. Data from a study of 386 violations between 2004 and 2022 found that in 34% of cases, prosecutors withheld evidence that did both at once, supporting innocence and undermining the case theory, and in 45% the suppressed material could have impeached a critical prosecution witness (per the Death Penalty Information Center's review of the research).

Giglio and Pitchess claims are different. They target officer credibility: prior dishonesty, false reports, biased conduct. In California, SB 1421 changed the access rules by opening certain records, including officer-involved shootings and findings of dishonesty, to public inspection, which can reduce the need for a Pitchess motion for those categories (as California practice guides note). For other misconduct categories, Pitchess remains the vehicle.

The timing point matters more than the taxonomy. Because the duty to disclose is ongoing, you can file at virtually any stage: after openings, mid-cross when the officer's testimony contradicts the bodycam, or even before sentencing (California practitioners report). The standard you must meet is a "reasonable probability" that the withheld evidence would have changed the outcome.

The practical problem is that 51% of successful Brady claims involve material that never reached the line prosecutor (the same study found), so you can't count on the state's file. You have to find the discrepancy yourself. A structured Brady list research workflow that cross-references police reports against bodycam footage and jail calls is how you surface the report-versus-video contradiction that becomes your mid-trial motion. Evidence Copilot's ingest-time comparison flags candidate inconsistencies across the discovery dump before anyone asks, with every flagged signal marked for attorney review.

Your checklist for each prosecution witness:

• What does the prosecution's file show that helps my client? That's Brady.
• What does this officer's history show about honesty? That's Giglio or Pitchess.
• Does my state have an SB 1421-style disclosure statute, or is a motion still required?
• Has anything new emerged since trial started? The duty to disclose hasn't stopped.

File the right motion under the right doctrine at the moment the evidence appears. That's the whole game.

The Discovery Review Workflow That Catches What Prosecutors Miss

Most Brady material doesn't announce itself. It hides in the gap between what the officer wrote at 2 a.m. and what the bodycam captured at 1:47, and nobody on the prosecution side is paid to notice the difference. That's your job, and a disciplined review workflow is how you do it before the preliminary hearing instead of after the verdict.

Start with the people, not the paper. In a study of 386 Brady violations between 2004 and 2022, suppressed evidence could have impeached a critical prosecution witness in 45% of cases, and 49% of violations occurred in homicide cases. So build your witness list first, then review everything through that lens: does this report, this call, this video touch the credibility of the person the State will actually put on the stand?

Then run the cross-reference in a fixed order:

• Police report against bodycam: compare the narrative timeline to the footage timestamp by timestamp, watching for details in the report that the camera doesn't support.
• 911 call against the report: the caller's words often diverge from how the responding officer later characterized the scene.
• Jail calls against witness statements: recorded conversations can undercut the account a cooperating witness gave.
• Prior statements against each other: an officer's preliminary-hearing testimony versus the incident report versus the footage.

The reason for the rigor is structural. In 51% of successful Brady claims, the material never reached the line prosecutor, a communication breakdown between law enforcement and the prosecution team that Professor McAward describes as "just a collection of parts that do not work well together." You can't count on the other side catching what their own intake process drops.

This is also where tooling earns its keep. When you ingest a discovery dump into Evidence Copilot, every new file is automatically compared against earlier evidence in the case, so a report-vs-bodycam conflict surfaces as an unreviewed signal before anyone thinks to ask. The one-click case brief then pulls the evidence inventory, key moments with citations to the exact second of video or page of the PDF, and flagged inconsistencies into a single grounded document your whole team can work from. For officer-credibility work specifically, the Brady list research workflow keeps that research attached to the case rather than scattered across tabs.

Two cautions from the ethics side. The NACDL's 2026 report, Parity in Practice, frames responsible AI adoption as an "ethical obligation" for defenders and warns against consumer chat apps because of confidentiality concerns, so keep review inside tools built for the work. And remember the materiality standard: you're building toward a "reasonable probability" that the evidence would change the outcome, which means every inconsistency you log should be tied to a quotable source you can put in front of a judge.

Using AI Ethically: What the NACDL's 2026 Report Means for Your Office

The NACDL's 2026 report draws a line that most defender offices haven't crossed yet, and the argument it makes is uncomfortable: if prosecutors are already using AI to review discovery faster, defenders have an ethical obligation to catch up, carefully.

Parity in Practice: The Defender's Duty to Ethically Use AI frames responsible AI adoption as a competency issue, not just an efficiency one. The report warns that falling behind creates a real disadvantage in practice, especially on cases where the discovery dump runs to hundreds of hours of bodycam and jail-call audio.

Where the report draws its sharpest line is tool selection. It flags "significant confidentiality worries" with commercial consumer AI products like ChatGPT and steers defenders toward enterprise or legal-specific platforms instead. That distinction matters more than it sounds.

A consumer chat window gives you fluent answers with no provenance. You paste a police report in, you get a summary back, and you have no way to verify which sentence produced which claim, and you've just uploaded client discovery to a general-purpose model. For a Brady violation argument, that's a problem: you need to show the court the exact moment in the bodycam where the officer's account diverges, not a paraphrase.

What a grounded workspace looks like in practice:

• Every AI answer cites the exact video second or PDF page it came from, so you can click through and verify before you cite it in a motion.
• All AI output is marked AI-generated and pending human review. The tool drafts; you decide.
• Originals stay immutable. Redactions and exports become derived files with a full audit trail.
• Auto-signals on ingest surface candidate inconsistencies (report vs. bodycam, statement vs. jail call) as unreviewed flags before you even ask a question.
• Pricing is published, not gated behind a sales call.

That last point matters for PD offices. Evidence Copilot lists Solo at $39/user/mo and PD Office at $750/attorney/yr, with a 7-day full-access trial and no card required. When you're building a budget case for a chief defender, published numbers beat a discovery call.

The NACDL's point stands regardless of which tool you pick. A Brady list research workflow that cross-references officer reports against bodycam and call audio catches the material that never reached the line prosecutor. That gap is where 51% of successful Brady claims live, according to the empirical research. Adopting AI responsibly isn't about keeping up with the Joneses across the aisle. It's about not missing the impeachment material sitting in your own discovery folder.