Looking beyond the label

In August, a taxidermy head bought for £50 at an antiques market in the South West was discovered not to be the badly stuffed fox it had been taken for, but an exceptionally rare thylacine, the extinct Tasmanian tiger. Once properly identified, it sold at auction in Bristol for £50,000. The story is an appealing parable about expertise, but also a warning about the confidence with which an inaccurate label can become accepted fact. Sometimes the most consequential professional skill is simply knowing when the description in front of us does not tell the whole story.

That problem recurs throughout this edition. Law depends upon records, classifications and formal frameworks. They give complex events structure and make consistent decisions possible. But every framework also directs attention: it makes some things visible and leaves others outside the frame. The solicitor’s task is therefore not only to apply the available material, but to recognise when it presents an incomplete picture.

Miroslava Marinova explores precisely this problem in legal research conducted with artificial intelligence. The most obvious danger is a fabricated authority. The subtler one is a genuine case placed in the wrong relationship to the wider law. A citation may exist, the vocabulary may be impeccable and the answer may read fluently, while an omitted qualification quietly changes the conclusion. Verifying that a case is real is necessary; establishing that the analysis is sound requires legal judgment.

The same distinction runs through Andrew Fei and Matthias Artzt’s examination of liability for agentic AI. The technology may plan and act with limited human intervention, but it cannot itself carry legal responsibility. Courts and regulators must look through the system to the people and institutions that designed, deployed, authorised or benefited from it. As autonomy increases, that attribution becomes harder, but the machine cannot become a convenient place for accountability to disappear.

Zoe Donohue finds a related evidential problem in clinical negligence. Medical records are rightly treated as powerful contemporaneous evidence, yet they are not a complete transcript of treatment. Notes may be rushed, selective or silent about a conversation that a patient remembers vividly. The absence of an entry is not necessarily evidence that an event did not occur, particularly where deficient record-keeping may itself have created the uncertainty. Respect for documentary evidence must therefore coexist with scrutiny of its limits.

Cleo Perry KC shows what can happen when the frame itself is too narrow. Child-protection systems developed primarily to examine harm within the family home can struggle to recognise exploitation operating through peer groups, criminal networks and online spaces. Language matters here: describing exploitation as a child’s “risky choice” does not merely record events badly; it transfers agency and blame to the person being harmed. Better protection begins with naming the reality correctly.

In commercial relationships, meanwhile, Anthony Vittadini and Claire Blackmore describe how operational reality can drift away from the signed contract. Verbal accommodations, tolerated underperformance and informal changes may gradually create a relationship bearing little resemblance to the agreement on file. Yet when a dispute crystallises, the formal machinery governing variation, waiver and relief becomes decisive. The lesson is not that the document is unreal, but that lawyers must understand both the document and the conduct surrounding it.

August also brought reports that permanent live facial-recognition cameras are planned for Oxford Street, comparing passers-by with police watchlists. It is a strikingly literal version of this edition’s underlying question: what happens when observation becomes identification, and identification becomes a basis for action? Whether assessing a face, a case citation, a medical note or a contractual record, professional judgment begins by resisting the temptation to mistake a match, label or omission for the whole truth.