Arrest Is Not Conviction: The Documented Harms of Publishing Names and Home Addresses

Crime data has a hard edge. The same public records that hold institutions accountable can ruin the lives of people who were never convicted of anything, and in a smaller but well-documented set of cases have ended those lives at the hands of vigilantes. This post is about what the record actually shows when names and home addresses get published alongside an arrest — and about the line we draw, deliberately, between publishing incidents and publishing people.

A note on framing. We are pro-transparency. We have spent the last several months on this blog documenting the harms of police departments going dark on address-level incident data and the importance of public records as a check on institutional power. None of that changes the fact that some categories of crime-adjacent publication produce documented harm to identifiable people who have not been adjudicated of anything. Both things can be true. The job of a credible crime data infrastructure is to know which is which.

The legal baseline

Arrest is not conviction. In US law, arrest establishes probable cause — a relatively low evidentiary bar — and triggers a process. Whether the arrest leads to a charge, whether the charge leads to a trial, and whether the trial leads to a conviction is a separate sequence of events that can take months or years and, in a non-trivial share of cases, produces an outcome other than conviction.

The US Bureau of Justice Statistics has consistently reported that a meaningful share of felony arrests do not result in conviction on the original charge — historically on the order of a third of state felony cases are dismissed, acquitted, or pled down to substantially different offenses. The fraction varies by jurisdiction and category and is a longstanding empirical fact. The implication for publishing practice is direct: a record of arrest is a record of an institutional event, not a finding about the person.

Three categories of harm follow when that distinction is ignored at the publication layer. We will take them one at a time.

Harm one: reputational and financial damage to the never-convicted

The most extensively documented category. The case most casual readers will know is Richard Jewell, the security guard who discovered the bomb at the Centennial Olympic Park during the 1996 Atlanta Olympics, alerted police, and helped evacuate the area. Three days later the Atlanta Journal-Constitution reported that he was under FBI investigation as a suspect, and within hours every major US news outlet had named him. Eighty-eight days later, the US Attorney's office formally cleared him. The actual bomber, Eric Rudolph, was identified years afterward and pled guilty in 2005. Jewell was never charged. The reputational damage he sustained — in particular the harassment of his family and the dismantling of his job prospects — is part of the public record because he and his lawyers documented it through subsequent litigation. He died in 2007 at 44.

The Jewell case is famous because the underlying event was a global news story. The same dynamic, at far smaller scale, repeats in arrest publishing every day. Until the legislative wave of the mid-2010s, an entire commercial industry — sites with names like Mugshots.com and BustedMugshots.com — scraped police booking photos by the millions, indexed them for search-engine visibility against the arrested person's name, and charged removal fees that ranged from $200 to several thousand dollars per image. The American Bar Association characterized the model as an online extortion scheme. The pattern of harm was consistent: an arrest that produced no conviction, no charges, or a dismissed case still produced a high-ranking Google result for the person's name that took years and money to remove, if it could be removed at all.

The legislative response has been concrete. By our count, at least eleven US states have passed laws restricting the practice. The chronological pattern:

• Georgia (2013-2014) — required free removal for cleared individuals
• Colorado (2014) — required free removal for the acquitted
• Connecticut (2014) — limited police disclosure of mugshots
• California (2015) — banned the solicitation of payment for removal
• Florida (2017) — mandated removal within ten days of a written request
• Arizona (2019) — prohibited removal fees with civil penalties up to $500 per day of non-compliance
• Missouri, South Carolina, Texas, Utah, and Virginia each enacted variants

The criminal cases caught up later. In May 2025, three of the four individuals behind Mugshots.com were convicted in California on charges that included attempted extortion and identity theft. The fact pattern in the conviction was essentially the fact pattern that critics had described for a decade: the site was monetizing the reputational damage it had itself manufactured by publication.

None of those eleven state laws would have been necessary if the underlying publication had been benign. The legislative record is, on its own terms, a documentation of the harm.

Harm two: vigilante violence enabled by public registries

This is the smaller category by case count but the more severe by individual outcome. Public sex offender registries, in their current form, are a US innovation of the 1990s — the federal Wetterling Act in 1994, Megan's Law in 1996, the Adam Walsh Act in 2006 — and they are explicit in publishing names, photographs, and home or work addresses. The policy theory was that public notification would deter recidivism. The empirical record on whether registries actually reduce sexual offense rates is, in the academic literature's own phrasing, mixed at best; the majority of published studies do not find statistically significant shifts in sexual offense trends following registry implementation.

What the record does show, with documented cases and dates, is registry-driven vigilante violence.

• Maine, April 2006.Stephen Marshall, age 20, used Maine's online sex offender registry to identify and locate registered individuals. He shot and killed two of them — William Elliott, 24, in Corinth, and Joseph Gray, 57, in Milo — before killing himself on a bus in Boston when police boarded it. Police recovered information he had pulled from the registry on twenty-nine people. Elliott's registry listing stemmed from a statutory offense as a 19-year-old involving a partner just under the age of consent — a category of registry entry that, in many states, is publicly listed indistinguishably from violent offenses. Marshall did not investigate; he picked names.
• Washington, June 2012. Patrick Drum killed two registered sex offenders, Gary Blanton (28) and Jerry Ray (56), in Clallam County after locating them through the state registry. He pleaded guilty and received two consecutive life sentences. He told investigators he intended to keep going.
• South Carolina, July 2013.Jeremy and Christine Moody killed Charles Marlin Parker, a registered sex offender, in his home in Union County. They also killed his wife Gretchen, who was not on the registry and was not the target of the registry lookup, because she was present when they arrived. They received life sentences. The case is a clean illustration of the registry's collateral-harm geometry: even when the listed individual is the intended target, bystanders are not protected by the listing.

These are not the only cases on the record, and they are not the only modes of harm. Academic surveys of registrants consistently document harassment, property damage, loss of employment, loss of housing, and threats against family members. Several research groups have proposed that the secondary effects of public listing — instability, isolation, employment exclusion — work against the reintegration outcomes most associated with reduced recidivism, which is the opposite of what the registries were designed to achieve.

The point is not that no one on a sex offender registry has committed a serious offense. Many have. The point is that public, name-and-address-level publication creates a permission structure for extra-judicial action by people who do not verify the underlying record, and the available case record documents that the permission structure has, repeatedly, been used.

Harm three: the exoneration that comes too late

The third category is what happens when the original arrest is later found to have been wrong on the merits, after the published record has done its work.

Brian Banks is the high-profile example most readers will recognize. Arrested at 16 in Long Beach in 2002 on a fabricated allegation, he accepted a plea bargain that resulted in approximately six years of incarceration and a placement on California's sex offender registry. He lost a USC football scholarship and the better part of a decade of his athletic career. The accuser admitted the fabrication in 2011. He was formally exonerated on May 24, 2012 — a decade after the arrest. He went on to play briefly in the NFL and is now a public advocate for wrongful-conviction reform.

Banks recovered some of what he lost because the case became national news and because he had unusual personal capacity to rebuild. Most exonerees do not. The National Registry of Exonerations, run jointly by the University of Michigan Law School and others, has tracked over 3,600 US exonerations since 1989. Each entry on that registry is a person whose name appeared on an arrest record, in many cases in a court docket and in news coverage, before the underlying basis collapsed. The cleanup is asymmetric: the original publication propagates fast and broadly; the exoneration does not.

This is the strongest argument for caution at the publication layer that does not depend on any contested empirical claim. Publication is fast and durable. Adjudication is slow and probabilistic. A system that prints names at the speed of arrest and corrects them at the speed of appellate review is going to leave a residue of harm on the people the system later determines should not have been there.

What this means for crime data infrastructure

The implication for how a crime data platform should be built falls out of the three harm categories above. We have written before on the three legal layers of crime data suppression; this post is the case for the secondary, voluntary layer that responsible publishers should apply on top of what law requires.

The line we draw, and the one we recommend to platform teams integrating crime data into consumer products, is between incidents and people.

• Incidents are publishable. An incident is an institutional event: a 911 call, a dispatch, a police report. Its publication is the basic transparency function we have been defending in our writing about the LAPD and other agencies that have gone dark. Time, type, and block-level location are the structured fields that make incident data useful for residents, real estate platforms, and analysts.
• Arrests with names are not. An arrest is a person-level event. Publishing it with a name and home address before adjudication creates the harm patterns documented above. The public-records statute may permit it; the responsibility question is independent.
• Mugshots are not. The booking photograph is the asset that the extortion industry was built on, and it is the asset whose publication produces the most persistent reputational damage relative to the institutional value it generates. The eleven states that have legislated against fee-for-removal mugshot publication did so for a reason.
• Registry-style address publication is the highest-risk case. The combination of name, home address, photograph, and a categorical label like “sex offender” in a queryable public interface is the exact configuration that the Marshall, Drum, and Moody cases documented as survivable to act on. We do not publish in this format, and we recommend that integrators not build interfaces that effectively reproduce it from other sources.

None of this is in tension with the transparency arguments we have made about the LAPD or about the importance of public access to incident data. Those arguments are about the institutional record of what happened in a place. This post is about the personal record of who was accused. They are different categories and they deserve different defaults.

The standard

A crime data platform that takes the documented harms seriously will publish structured incident data at the highest geographic resolution the source records sustain, will publish over time so that residents and analysts can see the trend in their own neighborhood, and will not publish personally identifying information about individuals who have not been adjudicated of anything. That is our standard. It is consistent with the transparency case for public records and with the historical evidence on the harm patterns described above.

Where the policy debate goes from here — particularly on the question of whether existing sex offender registries should be maintained in their current public form given the empirical record on effectiveness and the documented case record on vigilante harm — is a question for legislatures and courts, not for a data platform. But the data platform's own publishing posture is something it can choose. We have chosen ours, and the harms documented in this post are the reason.