Ad/Marketing Communication
This legal article/report forms part of my ongoing legal commentary on the use of artificial intelligence within the justice system. It supports my work in teaching, lecturing, and writing about AI and the law and is published to promote my practice. Not legal advice. Not Direct/Public Access. All instructions via clerks at Doughty Street Chambers. This legal article concerns AI Law
Artificial Intelligence (AI) is becoming ever more embedded in legal disputes, as demonstrated by three distinct judgments handed down early in 2025. Each case featured AI in a different way, and all three judgments drew upon questions that have long interested the courts: whether novel technologies fundamentally reshape established principles of law, or whether existing frameworks can simply be adapted.
Below, I briefly examine the three cases, their key facts, and how the role of AI proved significant enough to be cited in each decision. I will be going through each case in more detail shortly.
1. Getty Images (US) Inc & Ors v Stability AI Ltd, 2025 EWHC 38 (Ch)
https://www.judiciary.uk/judgments/getty-images-and-others-v-stability-ai/
Brief Facts:
Several Getty Images companies brought claims of copyright and related infringements against Stability AI, a company that develops the open-source image-generation platform “Stable Diffusion.” Getty alleged that Stability AI’s generative model was trained using millions of Getty-owned images without permission or licensing arrangements, leading to copyright infringement, database right infringement, trade mark infringement, and passing off.
How AI Played a Role:
Training Data and Copyright: Central to the case was the way Stability AI “scraped” images from the Getty Images platforms. Stability AI’s deep learning model, Stable Diffusion, relies on large datasets of images to produce new, synthetic outputs based on text or image prompts. Getty alleged that these AI-driven processes copied original works and thus infringed copyrights.
Representative Proceedings: While large portions of the ruling concern procedural questions (particularly the viability of a representative claim on behalf of many photographers and contributors), the key legal issue stems from how a generative AI model learns from copyrighted content. In citing AI, the court recognised the unique challenge of attributing liability when an algorithm is “trained” on vast volumes of material, much of which may be protected by intellectual property rights.
Key Takeaway:
The question of whether AI’s “training” process amounts to infringement of copyright attracted much attention. While the judgment here turned on procedural matters (especially around representative claims), the fact that AI underpinned all alleged infringements meant the court explicitly grappled with how to handle mass data scraping, the risk of recreating “substantial parts” of copyrighted works, and whether AI’s core methods differ in law from more conventional forms of infringement.
2. Unitel Direct Ltd v Racing Edge Auto Repairs Ltd & Ors 2025 EWCC 3 (CC)
https://www.bailii.org/ew/cases/Misc/2025/CC3.html
Brief Facts:
Multiple claims were brought by Unitel Direct Limited against various defendants for alleged unpaid fees under purported verbal business-to-business contracts for online advertising. The court dismissed the claims, partly due to uncertainty over whether valid contracts had been formed.
How AI Played a Role:
The claimant relied on transcripts of telephone conversations purportedly generated using a third-party AI-driven transcription service. The AI inserted headings and attempted to capture every spoken word.
The judge questioned the reliability of these AI-generated transcripts, noting inconsistencies, missing audio files, and a lack of suitable verification. Because the original audio was not produced and the transcription process remained opaque, the court gave limited weight to the AI-derived record.
Key Takeaway:
Although this was not an AI-related claim (it concerned contract formation and breach), AI was central to the evidence. Reliance on AI-driven transcripts introduced evidential uncertainties. The judge signalled that courts will demand robust foundations—full audio recordings, proven accuracy, consistent timestamps—before giving substantial weight to AI-generated documents. This highlights the growing need for standards that verify evidence produced by AI systems in litigation.
3. His Majesty’s Advocate v LM 2025 HCJAC 3
Brief Facts:
The respondent was convicted on multiple charges related to sexual offending, including grooming and rape of a young girl, as well as attempting to communicate indecently with another person he believed to be a child. The second “child” was in fact an adult decoy posing as a 14-year-old online.
How AI Played a Role:
The decoy used AI to edit her photograph, making her appear significantly younger. She then used this “underage” image to engage with the accused on social media, forming part of the evidence that he believed he was speaking to a minor.
The “paedophile hunter” group leveraged AI editing to bolster the realistic presentation of an underage decoy, thereby catching the accused’s predatory communications. The High Court referred directly to the AI’s role in facilitating the impersonation of a child, which underpinned the charges.
Key Takeaway:
This judgment reinforces how AI-manipulated images can be crucial evidence in criminal cases. Where an accused claims they did not know the other party was underage, AI tools that convincingly alter appearances can influence how the courts interpret entrapment, authenticity, and the accused’s knowledge or belief. For future criminal prosecutions involving AI-altered media, courts will likely scrutinise the technology’s reliability and the fairness of its use.
Looking Ahead: AI’s Legal Trajectory
Collectively, these three cases show that AI is no longer a peripheral concern but a recurring theme in litigation. From copyright infringement via training data, to the evidential reliability of transcripts, and to AI-altered photographs in covert investigations, these cases suggest several emergent trends:
- Training an AI system on copyrighted datasets is a legal minefield. Courts will likely face more conflicts over whether “transformational” AI outputs are permissible or whether the mere act of harvesting data (and potentially reproducing it) infringes existing rights.
- AI-assisted evidence (transcription, image manipulation, or otherwise) will face close judicial scrutiny.
- LM shows how readily AI can enhance or simulate an alternative reality—here, forging a lifelike younger appearance. Courts recognise both the investigative advantages and the possibility of entrapment or reliability arguments whenever AI-generated or AI-edited material is tendered.
- These disputes hint at the possibility of legislative or regulatory guidelines. Whether focusing on licensing and compliance for generative models or requiring expert certification for AI-evidence in criminal courts, lawmakers may soon need to fill gaps to keep pace with fast-evolving technology.
2025 will be an interesting year for AI in the law. These cases show that it has become intertwined in IP disputes, day-to-day commercial claims, and serious criminal prosecutions. All three decisions show that UK courts are developing a sophisticated understanding of AI, scrutinising it for reliability, authenticity, and compliance with established legal norms. I think we will see that trend intensify in the coming year.
