TALAMANA · THE AI LITERACY MAP FOR ARCHITECTURE AND DESIGN · Ethics & Provenance · AGE 16—18 · FACTUAL · CONTESTED

Whose work taught the machine

An image model learnt from people's pictures without asking them, and you work inside that fact.

The idea

An image model learned what a courtyard looks like from photographs people took and drawings people made — pictures of buildings, not the buildings themselves. For the large web-scraped datasets whose contents are documented, the people who made those pictures were not asked. For models whose training data is undisclosed, nobody outside can say. Whether the scraping was lawful is being decided, country by country, and the answers differ. Whether it was fair is a separate question. Whether a platform permits it is a third. Legal, ethical and allowed-by-the-platform are three different tests, and you work inside all three.

Why it matters

You cannot get outside this fact by choosing a different tool. You can only decide how you act inside it: what you prompt for, what you credit, what you refuse.

See it in the studio

A student types in the style of a living Indian architect and gets a convincing pastiche. In most copyright systems a style as such is not protected — particular works are. So the prompt may break no law. The question the card asks is different: could the student defend the act to that architect's face?

Watch for this

"It's on the internet, so it's free to use." Public is not the same as permitted. A photograph on a website is visible to everyone and still belongs to someone.

Try it

Pick a living photographer or architect whose work you admire. Generate three images "in their style." Then write the email you would send them explaining what you did. If you would not send it, you have your answer.

Prove it

Explain what it means, for the photographer, when a model can reproduce a living photographer's style — and state where you stand on prompts that name a living designer to borrow their style.

How it works

Courts have begun to rule, and the rulings are early and narrow. India · 24 July 2026 — ANI v OpenAI: interim ruling, not a final judgment. The Delhi High Court refused ANI's application for an interim injunction, treating the storage and use of ANI's articles for training as prima facie within India's fair-dealing exception on the facts before it, and finding that ANI had not shown substantial reproduction in the outputs it relied on. The suit continues. In the UK the High Court decided Getty Images v Stability AI in November 2025 on grounds specific to UK law and to where the training happened. The US Copyright Office published a report on training in 2025. None of this is settled, which is why the card is CONTESTED. Whatever the law decides, it will not tell you whether you could face the person whose work you borrowed.

What this idea builds on

What this idea opens up

Sources

Open this idea on the map · The complete map · Logika · RBDS AI Lab, India · revised every edition.

Age grows from 11 at the centre to 22 at the edge, and six sectors show the learning strands. Tab into the map and the arrow keys step from idea to idea, following the links where there is one. Enter opens the idea under the cursor, and E reads out its links and the reason recorded on each. Press slash for Search, question mark for the full key list, and Escape to leave. Open Ideas for the complete readable list, including what each idea builds on and what it opens up.

LOGIKA · RBDS AI LAB INDIA
ON-RAMP · AGE 11 · FIRST ENCOUNTERS, NOT GATES — IDEAS · — DEPENDENCIES
DONE
OPENS NEXT
SOLID — STANDS ON