The government is planning to use technology to resolve these delays. David Lammy, the Lord Chancellor, announced plans to trial virtual legal assistants in Crown Courts. These tools are meant to predict trial lengths and group similar hearings together to reduce administrative delays.
The current shift depends on proprietary software built by a small number of technology companies. These commercial platforms are closed systems. When the software that reads our laws and organizes our cases is private property, public accountability is lost. To protect citizens, we must build a public, decentralized AI network where algorithms are open, checked by peers, and trained on local data.
How delays are forcing courts to adopt technology
The Crown Court is a single, joint jurisdiction that faces severe pressures. These pressures include fewer court staff and more digital evidence in complex cases. The government has allocated a £2.785 billion budget for courts and tribunals to increase sitting days and run temporary "blitz courts". Even with this money, physical reforms may have a limited effect. An analysis by the Institute for Government suggests that restricting jury trials would reduce Crown Court time by less than 10%.
Delays explain why courts are looking at generative AI. This software can summarize documents and write contracts. Private clients are already using these systems. A freelance consultant used an automated firm named Garfield AI to file a claim in an English county court for £400. The automated system wrote the letters and pre-trial filings.
The chief executive of the Law Society, Ian Jeffery, warns that technology is not a replacement for court staff or funding. AI systems do not have moral reasoning. If courts use AI to replace lawyers or staff, they risk the fairness of the legal system.
Why closed systems fail in court
Closed generative AI tools have caused several errors in actual court cases. Large language models sometimes hallucinate, meaning they write false citations or invent legal precedents. In Australia, Chief Justice Stephen Gageler said judges now act as human filters to catch machine-generated mistakes. The Federal Court of Australia has introduced rules requiring lawyers to check all cited cases manually.
Why do these mistakes happen? Closed systems are trained on static data from the internet. They do not have local context. A study by Oxford and Potsdam universities showed that large commercial models change the meaning of text on sensitive political issues, including climate change and feminism. Even when told to keep the original meaning, the models adjusted drafts to fit specific biases. In a court system, these hidden biases can alter how a lawyer writes a brief or how a clerk summarizes a case.
The physical limits of computer networks
Running massive AI systems requires a large amount of electricity. Technology companies need huge data centers that put strain on national power grids. In several countries, local councils have blocked data center projects. These include projects in Virginia and London. A court strategy that sends every legal document to a central server uses a large amount of energy.
Centralized data collection also poses a risk to civil liberties. Security experts like Bruce Schneier and Jon Penney warn that AI-powered public surveillance can make citizens afraid to speak out or organize. If we combine public databases with corporate servers, we make it easier to monitor the population. Instead of sending data to central clouds, we can run legal algorithms on local devices. This keeps data private and reduces energy use.
How federated learning keeps data local
The legal sector can use federated learning to build public AI systems. In a federated framework, a central model is trained across a network of separate computers. These computers belong to courts and legal aid groups. The raw data never leaves the local computer. The software sends only mathematical updates to coordinate the model.
This setup helps organizations comply with privacy rules. For example, legal aid clinics must check for conflicts of interest before taking clients. If a clinic uploads a potential client's data to a commercial cloud tool, it can breach confidentiality. With a decentralized system, the clinic can query local, encrypted databases.
Federated learning allows models to learn from a wide range of legal situations. This prevents the bias that occurs when a single company controls the training data.
Open-source legal tools in action
Several open-source projects show that this technology is possible. A former lawyer at Latham & Watkins, William Chen, built a free tool named Mike. The tool runs document reviews and contract drafting. It is an alternative to expensive private platforms. Lawyers can download the entire code from GitHub and run it on their own computers. This ensures that sensitive files do not leave the office.
Another project is OpenJustice, developed by the Conflict Analytics Lab at Queen's University. OpenJustice uses specific legal rules to prevent hallucinations. This tool helps legal aid groups and law schools build public applications. It powers OpenCourt, a website that helps people understand if their employment layoffs are legal. Legal clinics also use OpenJustice to help low-income clients. This ensures that legal AI is used to defend public rights.
A plan for public legal technology
If the government relies on commercial software, the courts will depend on private monopolies. To prevent this, public organizations must stop buying closed AI platforms. Instead, public funds must go toward open-source tools and public databases. This will make courts faster while keeping the law accountable to the public.