The release of GPT-5 last month marked another step in the rapid development of generative artificial intelligence. Its arrival also provided the inspiration for my “5 Days, 5 Disputes” collection—a short series of articles, each examining an area of the law where AI is already colliding with established legal principles.
Each of the five articles in the series considers a different angle of how generative AI interacts with the law: the risks of false statements, the copyright questions raised by large-scale model training, the handling of confidential information, the potential for bias and the challenges of machine-driven contracting. View or download each article as a PDF document using the links below.
Day 1 – When AI Lies: Hallucinations and Falsehoods
AI tools sometimes “hallucinate”, inventing facts, advice, allegations or attributions that appear convincing but are wrong. When reputations or decisions are affected by these mistakes, they can become actionable wrongs. In part 1 of this series, I examine how defamation, negligent misstatement, misrepresentation and malicious falsehood apply to false AI outputs and what claimants need to do to prove publication, reliance and causation.
⬇️ View or download Day 1 as a PDF document
Day 2 – When AI Copies Too Closely: Copyright Disputes
Generative AI is trained on vast datasets, often including protected works. When its outputs reproduce portions of its training data or echo distinctive expressions of a work, copyright disputes may follow. In part 2 of this series, I explore how substantiality and unconscious copying principles may apply and what defences providers and users are likely to run. With the UK government having recently consulted on a copyright and AI framework and the Getty Images (US) Inc and others v Stability AI Ltd [2023] EWHC 3090 (Ch) judgment still pending, this area of the law appears on the verge of substantial change.
⬇️ View or download Day 2 as a PDF document
Day 3 – When AI Spills Secrets: Breach of Confidence
What happens when confidential business information finds its way into an AI model, whether through user prompts, insecure datasets or deliberate ingestion? In part 3, I analyse how the law on breach of confidence applies, what counts as “disclosure” in an AI output and what defences might be advanced.
⬇️ View or download Day 3 as a PDF document
Day 4 – When AI Tips the Scales: Bias and Unfairness
From recruitment and credit scoring to service allocation, AI-driven decision-making raises acute risks under the Equality Act 2010 and GDPR. While bias in training data is often behind discriminatory outcomes, proving causation and establishing comparators given the “black box” nature of many AI systems can be difficult. In part 4, I consider the litigation landscape and potential significance of SCHUFA (C634/21) to AI suppliers, developers and users, as well as what evidence might be key when pursuing or defending a claim.
⬇️ View or download Day 4 as a PDF document
Day 5 – When AI Signs on the Dotted Line: Contracts
Not only can AI draft contracts, it can also conclude them. From Alexa devices ordering goods to trading algorithms binding parties to deals, the question is not whether contracts can be formed, but who is bound and when. Drawing on cases from Thornton v Shoe Lane Parking Ltd [1971] 2 QB 163 to Quoine Pte Ltd v B2C2 Ltd [2020] SGCA 02, in the fifth and final part of this series I consider questions of authority, mistake and the incorporation of terms, and set out the evidence lawyers will need to field in order to resolve contract disputes in the AI age.
⬇️ View or download Day 5 as a PDF document
Final thoughts
Across the five areas explored in this series, a clear theme emerges: while the legal principles may be well established, they are being stretched and tested by the realities of AI. Courts are being asked to apply doctrines in contexts where the evidence is prompts, logs, models and datasets rather than paper files or witness testimony.
For clients and lawyers alike, the task is not just to know the law, but to understand how to prove (or resist) a claim in this new technical landscape. That is where careful analysis, technical fluency and strategic judgement will matter most.
