The UK Just Ordered Its Largest Court Database Deleted (And Your Legal Tech Startup Should Pay Attention)

The UK Ministry of Justice has ordered the deletion of Courtsdesk, the country's largest court reporting database. Within days, every record will be wiped. The platform that helped 1,500 journalists from 39 media outlets track criminal court cases is being shut down entirely.

The official reason? "Unauthorised sharing" of court information—specifically, that Courtsdesk sent data to a third-party AI company. But the story beneath the headline is more complicated, and it has direct implications for anyone building products that aggregate public data.

What Courtsdesk Actually Did

Courtsdesk was launched in 2020 with full government approval. The platform had an agreement with HM Courts & Tribunals Service, sign-off from the Lord Chancellor, and explicit approval from then-Justice Minister Chris Philp.

Its purpose was straightforward: make court information accessible to journalists. The UK's magistrates' courts were notoriously opaque. According to Courtsdesk's own data, the official court service's records were accurate just 4.2% of the time, and 1.6 million criminal hearings went ahead without any advance notice to the press.

"We built the only system that could tell journalists what was actually happening in the criminal courts," said Enda Leahy, Courtsdesk's founder.

That system worked. Reporters could search court lists and registers, track cases, and ensure that justice was being reported on. The alternative—calling individual courts, hoping for callbacks, manually checking listings—was so inefficient that most cases simply went uncovered.

Where It Went Wrong

In November, HMCTS issued a cessation notice citing the unauthorized sharing of court information with a third-party AI company. The specifics haven't been publicly detailed, but the implication is clear: Courtsdesk either fed court data into an AI system or shared it with a company that did.

Leahy says her company wrote to government agencies 16 times trying to save the service. They asked for the matter to be referred to the Information Commissioner's Office, but that request went nowhere. Even former minister Chris Philp personally approached current courts minister Sarah Sackman asking for the archive not to be deleted. The government refused.

An HMCTS spokesperson said journalists would continue to have "full access to court information to support accurate reporting." But given that Courtsdesk found two-thirds of courts regularly hearing cases without notifying journalists, that claim rings hollow.

The Open Justice Problem

The UK operates under a principle of "open justice"—criminal proceedings are meant to be public. In theory, anyone can walk into a courtroom and observe. In practice, the infrastructure for accessing that information remotely has always been inadequate.

Courtsdesk filled that gap. It wasn't creating new information; it was aggregating and organizing information that was already supposed to be public. The platform's value was in making the system work as intended.

Now that system is being dismantled, and there's no replacement. The Ministry of Justice's position seems to be that because court information is technically available—you can still show up in person, you can still request listings from individual courts—the deletion doesn't affect transparency.

That's technically true and practically absurd. Theoretical access to information that's functionally inaccessible isn't transparency.

The AI Trigger

What makes this case interesting for founders is the trigger. Courtsdesk didn't get shut down for being inaccurate or for violating privacy. It got shut down because court data touched an AI system.

We don't know the details. Maybe Courtsdesk was training models on court records. Maybe they were using an AI service for data processing that triggered contractual concerns. Maybe the Ministry of Justice simply decided that any AI involvement was a liability they didn't want to manage.

But the pattern here is troubling. Governments are increasingly treating AI integration as a categorical risk rather than evaluating specific harms. The response isn't "here are the rules for responsible AI use with public data"—it's "delete everything."

That approach creates a chilling effect. If building useful infrastructure on public data carries the risk of total shutdown when AI enters the picture, builders will either avoid AI integration entirely or operate in legal gray areas where they can't be monitored.

What This Means For Data Aggregators

If you're building a product that aggregates public data—court records, government filings, regulatory documents—this case should recalibrate your risk assessment.

The first lesson is that approval isn't permanent. Courtsdesk had explicit government sign-off from multiple officials. Six years later, a different government can decide that the same arrangement is unacceptable. Terms change, interpretations shift, and the political climate around AI is volatile enough that what's acceptable today may not be acceptable tomorrow.

The second lesson is about AI disclosure. If you're using any AI tools in your pipeline—even for processing, summarization, or internal analysis—you need to understand how that intersects with your data agreements. Many data access arrangements were written before AI was a mainstream concern. They may not explicitly prohibit AI use, but they may not authorize it either.

The third lesson is about data custody. Courtsdesk's archive is being deleted entirely. Years of organized court records, gone. If your product depends on institutional data access, consider what happens if that access gets revoked. Do you have the right to retain historical data? Can you continue operating with reduced scope? What's your fallback?

The Bigger Picture

This isn't really a story about one database. It's a story about the collision between open data principles and AI anxiety.

Public data has always been the foundation of accountability journalism, research, and civic engagement. The tools that make public data usable—aggregators, search engines, analysis platforms—serve the public interest even when they're built by private companies.

But AI is changing the calculus. When data gets fed into training sets, the downstream uses become unpredictable. Governments are responding by tightening controls on data that used to flow freely.

For founders, the strategic question is whether you can build defensible data infrastructure in this environment. The answer might be yes, but it requires more careful architecture around AI integration, clearer communication with data sources, and realistic contingency planning for access revocation.

Courtsdesk learned that lesson the hard way. Their database won't survive to teach it.