The Great AI Labelling Illusion; When Compliance Becomes More Important Than Trust

News is abound that from 2 August, organisations across Europe will proudly inform us that we’re chatting to AI. Websites will politely disclose that the smiling customer service agent is, in fact, a large language model, and marketing images will carry reassuring labels declaring they were generated by artificial intelligence.

Progress? Certainly. Protection? That’s another matter. I fear there is a real danger that this becomes AI’s equivalent of the cookie banner; renowned as a compliance exercise that consumes executive attention, legal budgets and countless steering committee meetings while leaving the far more important governance questions largely untouched.

We know this pattern. Tick the box. Pass the audit. Celebrate compliance. Meanwhile, nobody has asked whether the AI actually behaves reliably, whether it hallucinates under pressure, whether its recommendations drift over time or whether anyone is continuously assuring the authority that has been delegated to it.

A chatbot that begins with, ‘Hello, I’m an AI assistant’, may satisfy the regulator. As for the consumer, it tells almost nothing about whether the answer that follows is accurate, fair or safe.

It’s rather like requiring every aircraft to display a sticker reading, ‘This aeroplane is flown by autopilot’, while forgetting to ask whether anyone has inspected the engines.

Don’t get me wrong, yes, transparency is valuable. Consumers deserve to know when they’re interacting with AI rather than a human. However, transparency is the beginning of trust, not its destination.

The uncomfortable reality is that AI doesn’t become trustworthy because it announces itself or we label it so. It becomes trustworthy only when its behaviour is continuously tested, independently assured and governed by people who remain accountable for the consequences of delegating authority to machines.

Otherwise, we risk creating a generation of beautifully labelled AI systems that are impeccably compliant, wonderfully documented and meticulously audited … right up to the moment they make the wrong decision.

History suggests bureaucracy rarely fails because it asks too few questions. It usually fails because it asks the wrong ones.