An AI firm wins the first trial in England: Garfield AI charged £400 to claim £7,000

On June 22, 2026, The Guardian published a news story that marks a milestone in the history of Anglo-Saxon law: an artificial-intelligence-based law firm, Garfield AI, has won a case before an English court, in what is considered the first time a judicial proceeding has ended in victory thanks to the work of an artificial lawyer.

The protagonist of the case is Tamires Camal Taquidir, a freelance human resources consultant, who turned to Garfield AI to claim an unpaid debt of 7,000 pounds sterling. The total cost she paid the platform was approximately 400 pounds, a tiny fraction of what it would have cost to hire a traditional firm to litigate that same amount. Garfield AI prepared the legal claim letter, initiated the judicial proceedings and produced all the documentation necessary for the trial, which included four witness statements and a complete dossier of documents for a three-hour hearing held on May 14 at the Wandsworth County Court.

The case was not simple: the defendant, who hired solicitors (traditional lawyers) for their defense, filed a counterclaim with the apparent intention of raising the costs and pressure on Taquidir. The consultant openly acknowledged that the counterclaim 'sought to intimidate her,' but stated that she had accessible, cost-effective and competent support at all times. The court ruled in her favor and awarded her the money claimed.

It is essential to understand exactly what the AI did and what it did not do. Garfield AI carried out all the pre-trial legal work: case analysis, drafting of submissions, witness preparation and organization of the documentary file. However, the oral representation in the courtroom was handled by Dominic Li, a human barrister hired expressly for the hearing. Li told the Guardian that Garfield presented his client's case 'clearly and efficiently,' although he qualified that 'oral advocacy at trial remained essential and a fundamentally human exercise.' This distinction is key: the AI's merit lies in the preparation and instruction phase; the voice before the judge was human.

Garfield AI was authorized by the Solicitors Regulation Authority (SRA), the regulatory body for lawyers in England and Wales, in April 2025. Its business model is designed to cover claims of between 30 and 10,000 pounds, the segment known as 'small claims,' historically the great black hole of the judicial system: amounts too small to justify conventional legal fees, but significant enough to cause real harm to individuals and small businesses.

Philip Young, co-founder of Garfield AI, called the result a 'historic moment' for access to justice. His central argument is economic and social: many small businesses and self-employed people end up writing off uncollectible debts not because they lack a legal case, but because the cost of litigating exceeds what they can reasonably expect to recover. In that context, a platform that charges 400 pounds to manage a claim of 7,000 radically changes the calculation.

The article also contextualizes this success within a more turbulent landscape for AI in the legal world. The month before the ruling, the international firm Pinsent Masons voluntarily referred itself to the SRA after having misled a court on two occasions based on search results generated by an internal AI system. This type of incident—the so-called 'hallucinations' of language models, which generate nonexistent legal citations or facts with the appearance of truth—has shaken the British legal profession and fueled the debate over whether AI can be used safely in contexts where factual accuracy is literally a matter of justice.

The contrast between the two cases is illustrative: Pinsent Masons, a legal giant with ample resources, failed by relying too much on AI without sufficient oversight; Garfield AI, a regulated and specialized startup, succeeded with a model in which the AI does what it knows how to do well (process documents, structure arguments, draft submissions) and a human professional takes responsibility for the oral representation. The hybrid model, at least in this case, worked.

From the perspective of agentic artificial intelligence, the Garfield AI case is especially relevant. We are not dealing with a chatbot that answers legal questions or an assistant that suggests contractual clauses: we are dealing with a system that acts as an autonomous legal agent, makes decisions about procedural strategy, drafts documents with real legal effects and coordinates interaction with the judicial system. That this agent has acted within a regulated framework—with express authorization from the SRA—and with a human at the critical point of oral representation suggests a governance model that could be a reference for other jurisdictions.

In short: an HR consultant paid 400 pounds to an AI, recovered 7,000 pounds before an English judge, and the legal profession has spent weeks digesting the implications. The question left hanging is not whether AI can win trials—it already has—but how fast this model scales, what regulatory boundaries will shift, and what happens when AI not only prepares the case, but also argues in the courtroom.