iPhone & iPad — two quick taps:
A 7.7-terabyte LAPD leak. A wrongly-jailed Tennessee grandmother. 1.4 million ICE database queries hidden under a sanctuary law. A federal misconduct database deleted on Day 1 of a new administration. Thirteen states quietly auto-sealing arrests. The American arrest record is being rebuilt in real time — and the rebuild is going in two opposite directions at once.
For two decades, the direction of travel on American arrest records was one way: more access, more aggregation, more permanence. Mugshots got indexed by Google. Jail rosters got APIs. Background-check companies bought arrest data in bulk. A 1985 arrest that ended in dismissal would still surface in a tenant screening forty years later.
In 2026, four force vectors are pulling in four different directions at the same time, and the net effect is something nobody planned. State legislatures are auto-sealing arrests at scale. Federal accountability infrastructure for police misconduct has been deactivated. Local sanctuary laws are being tested by lawsuits over data quietly shared with ICE. And a 7.7-terabyte LAPD breach in April 2026 dumped officers' personnel files onto the dark web, where they remain.
Thirteen states and the District of Columbia have now passed laws that meet the Clean Slate Initiative's automatic-sealing criteria — with Illinois becoming the 13th state in 2024 when Governor Pritzker signed Public Act 104-0459 (Clean Slate Initiative) (Governor's Office). The Clean Slate Initiative's own states-map confirms the count: "thirteen states, and Washington D.C., have passed laws that meet our criteria" (cleanslateinitiative.org/states). Virginia's new record-sealing statute takes effect July 1, 2026, opening "an entirely new practice area for attorneys across the Commonwealth" according to the Virginia State Bar (Virginia State Crime Commission) (Virginia State Bar).
Pennsylvania's Clean Slate law — the first in the country, signed 2018 — had cleared nearly 35 million records within its first year, per the Center for American Progress' release marking the anniversary (Center for American Progress). By the program's second phase, the Pennsylvania House confirmed the rolling automation continues (PA House). Michigan's automatic-expungement system, which came online April 11, 2023, is described by the state Attorney General as the second-generation Clean Slate model — application-free, MSP-database-driven (Michigan AG) (Michigan Legal Help).
The Clean Slate Initiative's threshold matters because each of these laws includes arrest records, not just convictions. An arrest without conviction, in a Clean Slate state, is now systematically being walled off from the commercial record. That is a quiet but real reversal of the 2005–2020 trend.
On January 20, 2025, President Trump signed Executive Order 14148 rescinding Biden's Executive Order 14074 on policing. The U.S. Department of Justice announced the same week that the NLEAD "is no longer active" and had been "decommissioned in accordance with federal standards" (Congressional Research Service).
What was in NLEAD before it went dark, per the Bureau of Justice Statistics (BJS)’ final report covering 2018–2023: 4,790 qualifying incidents of officer misconduct involving 4,011 federal law enforcement officers across 94 agencies, drawn from roughly 148,000 federal officers. Of those incidents, 63% (3,031) were sustained complaints or disciplinary records. The breakdown: 737 terminations for misconduct, 481 resignations or retirements under misconduct investigation, 311 criminal convictions, 230 suspensions of enforcement authorities (decertifications) (BJS — NLEAD 2018–2023) (BJS report PDF).
The Appeal's analysis of the underlying dataset before deletion found the records were concentrated in Customs and Border Protection (CBP) and Bureau of Prisons (BOP) — the border and the prison system. NPR reported that the National Decertification Index (NDI), run by the International Association of Directors of Law Enforcement Standards and Training (IADLEST) membership group of state regulators, survives because it sits outside the federal executive branch (The Appeal) (NPR).
NLEAD's deletion does not delete the underlying conduct. It removes the federal aggregation layer. A CBP officer with a sustained excessive-force finding who transfers to a state agency now passes through state-level background screens that no longer query a federal misconduct database. The conduct exists; the registry that surfaced it does not.
Brady (1963) and Giglio (1972) require prosecutors to disclose to defendants any material that could impeach a testifying officer's credibility. For sixty years, those disclosure files lived in prosecutor desks. In 2026 they are moving onto the public web.
A national-scope public-facing aggregator — giglio-bradylist.com — compiles officer-credibility records sourced from across U.S. prosecutorial offices, including police-misconduct records, public complaints, and use-of-force reports (Brady List database), with individual officer profiles searchable (individual search). The site is operated by a private aggregator; coverage and accuracy vary by jurisdiction.
State legislatures are catching up. Minnesota's SF 2838 proposes statutory disclosure of peace-officer personnel data tied to Brady/Giglio findings (Minnesota Revisor — SF 2838). New Jersey's Attorney General Directive 2019-6 already mandates that prosecutors maintain and share Brady/Giglio lists across agencies; a 2025 appellate brief documents the operational architecture (NJ Courts — A-0223-25).
The implication for arrest records: when an officer with a Brady designation makes an arrest, that arrest now travels with a paper trail the defendant can find without subpoena. For employers and tenant-screeners querying the same officer's name in a background check context, the result is unprecedented public scrutiny — but only of those officers whose prosecutors have published. The state-by-state variance is enormous.
By 2026, body-worn cameras are standard issue for roughly 80% of large U.S. police departments. The recording of an arrest, in most jurisdictions, exists. Whether the public — or the arrestee — can obtain a copy on a useful timeline is a different question, governed by a patchwork of state release laws, "critical-incident" protocols, and Freedom of Information Act (FOIA)-style exceptions.
California's Senate Bill 1421 (2019) and AB 748 require release of body-cam video involving officer use of force within 45 days. New York's police-disclosure repeal of Civil Rights Law §50-a (2020) opened officer disciplinary records but left the per-agency release calendar uneven. Texas operates under a "critical-incident video" framework that allows departments to wait until investigation is complete — which can mean years.
The raw, unedited file — the actual evidence — is held by the agency on storage infrastructure that was largely not designed for the volumes involved. Departments are now contracting with cloud-storage vendors whose retention defaults sometimes overwrite footage at 90 days unless flagged. An arrestee who doesn't file a preservation letter within that window may lose access permanently.
On May 5, 2026, the Rural Organizing Project filed suit against the Oregon State Police, alleging years of state-police data-sharing with federal immigration agencies in violation of Oregon's sanctuary statute. ROP's complaint states that "federal immigration agencies have exploited access to Oregonians' information 1.4 million times in the past year" (Rural Organizing Project — May 5, 2026 lawsuit announcement). The Oregon Public Broadcasting reporting frames the volume: state police allowed two databases to be queried by ICE for years (OPB — May 6, 2026).
ICE's own enforcement footprint, per Transactional Records Access Clearinghouse (TRAC) at Syracuse: 32,531 people booked into detention by ICE during March 2026, of whom 29,606 were arrested by ICE and 2,925 by CBP (TRAC Immigration). The TRAC running comparison shows the month-over-month decline from 39,694 in January 2026 (TRAC — What's New).
The detention population reached a record. The American Immigration Council documented "a record 73,000 people detained as of mid-January" 2026 — an increase of more than 75% in a single year (American Immigration Council). Vera Institute's dashboard adds the facility-side metric: "ICE has opened 152 new detention facilities during the second Trump" term, and "reopened 170 facilities that it had not been using in the year prior" (Vera Institute). The ACLU's parallel reporting documents the conditions (ACLU — Deaths in Detention).
The composition has shifted as well. Where 2024 ICE bookings skewed toward people with criminal records, 2026 bookings increasingly do not — a trend the Vera dashboard surfaces and TRAC's longitudinal data confirms.
On July 14, Angela Lipps — a 50-year-old grandmother babysitting in Tennessee — was arrested at gunpoint by U.S. Marshals on a warrant generated by a facial-recognition match the West Fargo, North Dakota police had run through Clearview AI and shared with Fargo police, who executed the arrest (CNN — March 29, 2026). The actual suspect in the Fargo fraud case bore a passing resemblance. The match was wrong.
Lipps spent nearly six months in custody — first in Tennessee, then transferred to North Dakota — before being released on Christmas Eve (InForum / Fargo) (Privacy Guides). The ACLU's running tally of facial-recognition false arrests now lists Lipps alongside Robert Williams, Michael Oliver, Nijeer Parks, and at least a dozen others (ACLU — wrongful arrests dossier).
The arrest record itself is the part that does not automatically self-correct. Lipps' booking photo, her name in Tennessee and North Dakota dockets, and the warrant entry persist in commercial databases that scraped them within hours of the booking. Expungement is statutorily available in both states, but it requires action — a petition, in most counties — and it does not retroactively reach the third-party brokers who already indexed the entry.
An arrest record generated by a false-positive AI match is statistically indistinguishable, in commercial-broker databases, from one generated by probable cause. The legal-error layer is invisible to the downstream consumer. Until Chatrie v. United States or a similar Fourth Amendment vehicle gives the Supreme Court an opportunity to address the constitutional baseline, the cleanup remains entirely on the wrongly-arrested individual.
Chatrie v. United States is set for oral argument before the Supreme Court on April 27, 2026 — addressing whether the geofence warrant in that case violated the Fourth Amendment (Knight First Amendment Institute) (Brennan Center). The Brookings analysis previews how the justices may rule (Brookings). Geofence warrants are not the same instrument as facial-recognition matches, but the underlying question — what counts as particularized probable cause when the search is statistical — is the same constitutional terrain.
The American jail roster is a county-by-county artifact, not a federal one. There are roughly 3,000 county jails, each with its own access norms, software vendor, and policy on whether the daily booking log gets posted online. For most of the 2010s, the direction of travel was toward more posting — sheriffs liked the transparency optics, and vendors charged less for public-facing modules.
In 2026, several counties have pulled rosters back. Some cite the LAPD-style breach risk: a jail roster published in real time is also a jail roster harvestable by anyone, including extortion operators. Others cite arrestee-privacy concerns post-Clean Slate. Florida's 901.43 statute explicitly treats commercial mugshot publication as an "unfair or deceptive trade practice" — and the 2021 amendments expanded the statute's reach to entities whose "primary business model is the publishing and disseminating of arrest booking photographs for a commercial purpose or pecuniary gain" (Florida Statutes Ch. 901.43) (FindLaw — FL §901.43).
North Carolina is in motion: House Bill 1186 directs the North Carolina Collaboratory to establish a pilot program (NC General Assembly — H.B. 1186), and a compromise jail-mugshot bill passed the NC House in 2026 allowing mugshot removal from for-profit websites when charges are dismissed or the person is found not guilty (NC Newsline). The Pennsylvania General Assembly's own co-sponsorship memo notes "over 16 states have laws that prohibit these websites", framing the activity as "nothing more than extortion" (PA General Assembly memo). The American Bar Association tracks the state-by-state expansion (ABA Journal) (Wikipedia — Mug-shot publishing).
Maryland is now poised to join. Governor Wes Moore's office signed the first major bill-signing of 2025 covering education and veterans bills (Maryland Governor's Office). The 2026 session ended with bills covering affordability and criminal-justice measures (CNS Maryland) (Maryland Matters), with HB 360 — Maryland's Clean Slate Act of 2026 — analyzed in detail in the Department of Legislative Services fiscal note: it "will likely increase the overall number of expungements for various misdemeanors and felonies" (Maryland DLS — HB 360 fiscal note) (Clean Slate testimony).
The breach itself occurred on March 20, 2026, when the extortion group WorldLeaks obtained access through a vulnerability in a system used by the L.A. city attorney's office and posted the stolen data; the Los Angeles Times surfaced the story publicly on April 7, 2026, reporting that 7.7 terabytes of sensitive LAPD records had been compromised: "Leaked materials contain discovery documents" along with officer files (Los Angeles Times — April 7). The April 10 follow-up by the same paper explained the attack vector: "The hacking group appears to have exploited vulnerabilities in a system used by the Los Angeles city" (Los Angeles Times — April 10). A third LA Times piece on April 22 detailed how city officials struggled to explain the breach to the public (Los Angeles Times — April 22). Privacy Wire and SANS independently documented that the extortion gang WorldLeaks stole 7.7 terabytes and 337,000 files (Privacy Wire) (SANS NewsBites).
The 2024 National Public Data breach remains the larger-aperture event. According to the lawsuit, the USDoD cybercrime group claimed personal data on 2.9 billion people (Security Management / ASIS) (University of Hawaii — Info Security). Congressman Ritchie Torres released a Congressional investigative report on the breach (Rep. Ritchie Torres). The data set included Social Security numbers, full names, addresses, phone numbers, and email addresses linked to 137 million unique email addresses. The arrest-record subset of NPD was small in absolute terms but disproportionately damaging: arrest records keyed to live Social Security numbers are exactly what identity-theft operators want.
First: check your state. If you live in a Clean Slate state — Pennsylvania, Michigan, Utah, Connecticut, Virginia (after July 1, 2026), Delaware, New Jersey, Oklahoma, Colorado, California, New York, Illinois, Minnesota, or D.C. — your eligible arrest may be auto-sealed without action (Clean Slate Initiative — states map). If you live in Florida, file under §901.43 with the publishing entity in writing; refusal to remove constitutes a deceptive trade practice (Fla. Stat. 901.43). In North Carolina, watch H.B. 1186 and the compromise jail-mugshot bill (NC Newsline).
Start with the originating jurisdiction's court records portal — most counties now expose case-search APIs. For federal arrest events, Public Access to Court Electronic Records (PACER) and the federal-court Case Management/Electronic Case Files (CM/ECF) remain the source of truth. For police-misconduct context, Brady/Giglio lists are increasingly indexed (brady-list database). Cross-reference with the ACLU's wrongful-arrest dossier where AI-driven arrests are at issue (ACLU).
The Fair Credit Reporting Act (FCRA) still controls. After NLEAD's deletion, federal-officer screening lost a layer; verify with the National Decertification Index (run by IADLEST, not deleted) (NPR — NDI explainer). Be aware that Clean Slate states' automatic-sealing means an arrest that did not result in conviction may not appear, and querying it may violate state law.
Use BJS’ final NLEAD report for the historical federal misconduct dataset (BJS NLEAD 2018–2023). For ICE-detention monitoring, TRAC's monthly bookings dataset is the authoritative source (TRAC Immigration); cross-validate with Vera Institute's facility-count tracker (Vera Institute). For Brady/Giglio enforcement, the New Jersey AG Directive 2019-6 docket provides the operational template (NJ Courts brief).
Federal Clean Slate Act (H.R. 3114, 119th Congress). Introduced in the spring of 2025, the federal Clean Slate Act would require automatic sealing of certain federal criminal records. As of late May 2026, the bill remains in committee. The political math at this writing is unfavorable for clean federal passage, but elements may move through appropriations riders or omnibus crime bills.
Chatrie v. United States. Oral argument set for April 27, 2026 at the Supreme Court; decision likely by late June 2026. The geofence-warrant question is the closest 2026 SCOTUS vehicle for the broader question of statistical-evidence-driven arrests (Knight Institute case page) (Brennan Center).
Utah S.B. 153. The voter-privacy bill that changes how Utah voter-registration records are managed has drawn organized opposition and active disclosure to affected voters (Utah Legislature SB 153 — HTML) (SB 153 — full PDF). Its effective-date provisions are governed by the bill text as enacted; the bill itself is on the Utah Legislature's 2026 session page.
Maryland HB 360. Maryland's Clean Slate Act of 2026 cleared the Department of Legislative Services fiscal-note review and was teed up for action at the Governor's bill-signing series after sine die (MD DLS fiscal note) (MD Clean Slate testimony). Whether it makes Maryland the fourteenth Clean Slate state turns on Governor Moore's signature in the final bill-signing rounds (Maryland Matters).
The thirteen-state Clean Slate map, the deleted NLEAD, the growing Brady/Giglio public databases, the ICE detention expansion, the LAPD breach, and the Lipps case all stand at the same moment in the same calendar. They are not policy events that share a sponsor. They share a result: the arrest record itself, as a data object, is being rebuilt — and the rebuild is going in two directions at once.