Forum

Transparency Is the Hinge: A Week of Votes and a Trial in London

The dark glass facade of a building beneath a blue sky with large white clouds

Two very different institutions in London spent the past fortnight on the same underlying question: can anyone know what an AI model was trained on?

In Parliament, the Data (Use and Access) Bill has become the stage for a long standoff. On 2 June the House of Lords voted 242 to 116 to add a clause, moved by Baroness Kidron, that would require AI developers to be transparent about the copyrighted material used to train their models. The next day the House of Commons removed it again, by 317 votes to 185. It is not the first time the two Houses have traded amendments on this point, and the bill’s final shape is still unsettled.

On 9 June, at the High Court in London, Getty Images and Stability AI began a trial expected to last about three weeks. Getty alleges that Stability used millions of its images without permission to train Stable Diffusion. Stability denies the claims. It is the first case of its kind to reach trial in the English courts.

Why transparency keeps coming back

The parliamentary fight can look like procedure, but its substance is simple. Creative industries argue that without disclosure of training data, no licensing market can function and no enforcement is possible, because rights holders cannot see what was used. The government has argued against putting that duty into this bill, while promising further work on AI and copyright.

The trial shows why the argument matters. Litigation over training is hard partly because the evidence sits inside companies, on servers that may be in other countries, and in datasets that change over time. A court can order disclosure in a specific case. A transparency rule would make some of that information available before anyone has to sue.

A design perspective on disclosure

Designers know that the hardest part of any system is making the invisible legible. Transparency in AI is that problem at an industrial scale. Useful disclosure would need to answer questions a creator can act on: Was my portfolio site crawled? Was my work licensed from an agency? Was it excluded after I objected?

That is a higher bar than a long list of URLs. It suggests that the debate in Westminster is not only legal but also a matter of information design: what form a disclosure should take so that an illustrator, a photographer or a type designer can actually read it.

What creators can do now

Neither the vote nor the verdict will settle things this month. In the meantime, there are steps worth taking.

  • Know where your work lives. Keep a record of the platforms, agencies and sites that host your images, and the terms under which they do so.
  • Use the signals available. Crawler controls and machine-readable rights statements are imperfect, but they create a record of what you asked for.
  • Document your own AI use as carefully as you would want others to document theirs. If you expect transparency from model developers, practice it in your own studio.
  • Follow the detail. Much of what matters, such as which claims survive and which jurisdiction applies, will emerge from the trial’s daily record rather than its final headline.

Transparency inside competitions

Awards are a small model of the same problem. A jury evaluating AI-assisted work needs to know what the entrant did and what the machine did. The 3rd International AI Award 2025 builds that into entry: participants describe their models, prompts and ethical considerations, and confirm they hold the rights to what they submit. Judging begins on 10 June, with the final interviews scheduled for 15 July.

None of that replaces public policy. It does show that disclosure can be routine rather than adversarial when it is designed into a process from the start.

The question to keep asking

Whatever happens next in Parliament or in court, the most useful question for creative professionals is not whether AI training is good or bad. It is whether the people whose work is involved can see what is happening to it. Every week that question stays unanswered, trust erodes on both sides. Transparency is the hinge on which licensing, fair compensation and better tools all turn.