Projects · Cross-cutting

Intellectual property and AI.

A side project that runs through all the others. This site was built with AI, the laws project will vectorise texts, golf will hand playing data to an AI coach: each time, the question is who owns what. This page opens the discussion.

The questions

What AI shakes up.

  • Who owns what AI produces?

    In Europe as in the United States, copyright protects an original human creation. Content generated without real human input may belong to no one. Where is the line when a human chooses, corrects and assembles?

  • Training data

    Models learn from existing works. EU law allows text and data mining unless rights holders opt out. How is that opt-out expressed, and who checks it?

  • Code written with an assistant

    An assistant can reproduce fragments of open-source code. Who is responsible, and how do you spot it before publishing?

  • Inventions and patents

    Patent offices have refused to name an AI as inventor. What becomes of an invention found with the help of AI?

  • Transparency

    The EU AI Act requires flagging some generated content and obliges providers of large models to document their training data. What must be stated, and where?

  • Providers’ terms

    AI vendors generally assign outputs to the user. That guarantees neither that the outputs can be protected nor that they are free of third-party rights.

Cross-cutting

What it touches in each project.

  • This website

    Code, design and part of the texts produced with Claude Code: what is really mine?

  • French laws under scrutiny

    Official texts are free, but the databases that distribute them have their own licences. What can be vectorised and published?

  • Finance

    Prices come from third-party providers with their own reuse terms. How far can they be stored and displayed?

  • Golf

    Scores, location and playing habits shared with friends and handed to an AI coach: who owns this data?

  • Home automation

    Home data is personal: handing it to an AI model means knowing where it goes.

  • AI lab

    Every experiment raises the question of the data used and of the rights over what comes out.

Case study

This site, built with Claude.

What comes from me

The project, the choices, the order of the sections, the feedback at every step, the final selection and the decision to publish. My photograph, the source of the portrait.

What comes from the AI

Writing the code, layout proposals, part of the texts and diagrams, the script that computes the portrait.

A first hypothesis, to be checked: it is the human input (choices, selection, arrangement) that can ground protection, not the raw output.Hence the value of keeping a record of those choices, which the Git history does, and of saying so, which the legal notice does.

The approach

From texts to a personal charter.

  1. 01

    Map the texts

    Luxembourg and French copyright law, the 2019 EU copyright directive, the EU AI Act, patent law.

  2. 02

    Read providers’ terms

    Compare what AI vendors say about outputs, data sent and its reuse.

  3. 03

    Follow the rulings

    Court rulings, positions of patent offices and authorities, in Europe and the United States.

  4. 04

    Apply to the projects

    Go through each project on the site with these markers and note what raises questions.

  5. 05

    Write a usage charter

    Draw simple rules for my own projects: what I disclose, what I check, what I rule out.

Caveats

A reflection, not legal advice.

  • Not legal advice. This page gathers questions and leads. For a real case, consult a lawyer or an intellectual property attorney.
  • Fast-moving law. Texts, rulings and providers’ terms change every year: every finding will be dated.
  • Answers that vary by country. Luxembourg, France, the European Union and the United States do not always answer the same way.