Law & Policy

Court Records Are Public By Design, and Searchable By Accident

August 1, 2026 10 min read Haven Team

Someone sues their landlord over a deposit and wins. Two years later, a rental application is declined and nobody can quite say why. The case is public, as court cases are supposed to be. What changed is that reading it no longer requires anyone to drive to a courthouse and ask a clerk for a box.


Open justice is one of the older principles in common-law systems, and the reasoning behind it is sound. Proceedings that the public can observe are harder to conduct improperly. The US Supreme Court has recognised a common-law right to inspect and copy judicial records, and a constitutional right of access to criminal proceedings. Similar principles hold across most democracies in some form.

That framework was built when a public record meant a document held at a specific counter in a specific building during specific hours. Reading one took a trip. Reading a thousand about a thousand different people took a career. The friction was never a legal protection, and it was doing an enormous amount of protective work anyway.

Practical obscurity has a name because a court noticed it

The doctrine was articulated clearly in 1989, in a case about whether the FBI had to release a compiled criminal history under freedom of information law. Every individual item in the file was drawn from public records held by courts and police departments. The Supreme Court held that the compilation was different in kind from the scattered originals, and that a privacy interest attached to the aggregate that did not attach to the parts.

The same facts, gathered from many places into one place, become a new thing. Whether they were individually public does not settle what the compilation is.

That reasoning has aged well as a description and poorly as a protection. Bulk electronic access to dockets means the compilation now happens automatically, continuously, and by parties with a commercial interest in it. The rule from 1989 constrains what one government agency must hand over. It does not constrain a scraper.

What a filing actually contains

Court rules do require redaction of certain fields. In US federal civil practice, the relevant rule limits social security and taxpayer numbers to the last four digits, financial account numbers likewise, refers to minors by initials, and gives only the year of birth. The criminal rules mirror this.

Two things about that list matter more than its contents. The first is that the obligation sits with the filer, not the court. A clerk does not check. The second is that the list is short, and everything outside it is fair game:

Detail Typically redacted by rule?
Full social security number Yes, last four only
Full date of birth Yes, year only
Name of a party No
Home address on the caption or in an affidavit Frequently not
Employer, salary, medical history in a filed exhibit No, unless separately sealed
Names of family members in the narrative No
Allegations later withdrawn or disproved No, the filing stays on the docket

That last row is where much of the harm concentrates. A docket is a record of what was filed, not a record of what turned out to be true. A dismissed claim, a withdrawn accusation, an emergency motion the court denied within a week: all of it is permanently visible, in the words of the party who wanted it believed, with the outcome recorded in a separate entry that a downstream reader may never reach.

Redaction that is not redaction

Even where a filer intends to redact, the mechanics defeat them regularly. A black rectangle drawn over text in a PDF viewer is a graphic placed above a text layer that remains fully present in the file. Copying the region, or extracting the text stream, returns the original.

This has surfaced in high-profile filings often enough to be a genre, and the same failure mode reaches ordinary cases where nobody is watching closely enough to notice. Our guide to redaction failures covers the mechanism, including the related case of pixelated or blurred images that can be inverted when the underlying content is drawn from a small set.

If you are filing something

Redact in a tool that removes the underlying content rather than covering it, then verify by extracting the text from the finished file and reading what comes out. Print-to-image before filing is a blunt method that works. Checking is the step, not the intention to redact.

Where the records go next

Court data does not stay at the court. Bulk access feeds a downstream industry, and the products built on it reach people who never consult a docket:

The pattern is the same one that makes voter files and credit header data valuable to brokers. A record created for one civic purpose becomes an input to a commercial one, and the transfer requires no permission because the original was public.

Europe took a different route

Civil-law jurisdictions have generally treated the publication of judgments as separable from the openness of proceedings. Many publish decisions with the names of natural persons replaced by initials or codes as a matter of routine, and the Court of Justice of the European Union moved to anonymise the names of individuals in referred cases as standard practice.

The reasoning is that the public interest is in the reasoning of the court, which is what makes the law knowable, and that identifying the individual serves it only rarely. Common-law systems have historically weighted named parties more heavily, on the view that anonymous justice invites the abuses open justice exists to prevent.

Both positions are defensible. The point worth taking is that the balance was struck deliberately in one tradition and left to inherit the effects of digitisation in the other. Delisting under European law can remove a search result pointing to a report of a case, and does not touch the register itself, which is a narrower remedy than it is often described as.

What is available to an individual

Less than there should be, and more than nothing.

The structural fix is not an individual one. It is court systems distinguishing between access and bulk republication, adopting default pseudonymisation for categories where naming serves no public interest, and treating outcome data as part of the record rather than an appendix. Several US state judiciaries have moved in that direction. The default in most places remains a docket built for a filing cabinet and served over the open web.

Try Haven free for 15 days

Encrypted email and chat in one app. No credit card required.

Get Started →