If a text or an image was made entirely by AI and all you did was type one line of instructions, copyright protection is doubtful, and in practice you may be unable to stop someone copying it. If you reviewed, chose and rewrote it, the part that is yours is protected. In Italy the law says so from October 10, 2025; in the United States the Copyright Office says so. Here is what it means for a business that publishes catalogs, newsletters and posts.
Venice, 1469: the first printing privilege
You'll say: Professor, we start from printing to talk about ChatGPT? Yes, because the question is the same.
In 1469 a German, Johannes de Spira (Johann of Speyer), opens one of Italy's first print shops in Venice. According to the Wikipedia entry on Johannes de Spira, the Venetian Senate granted him a printing monopoly for five years. His first book was an edition of Cicero's Epistolae ad familiares, in 100 copies. Then he died of the plague while printing Augustine's City of God; Wendelin finished the work and ran the business until 1477.
Let me be precise about what I know and don't: I found no source saying for certain what happened to the privilege after his death, so I won't tell that part. What matters is the mechanism. A new technology made copying a text very easy, and the city answered with a legal tool: protection for whoever invests and takes the risk. Cicero's text belonged to everyone; what was protected was the work of turning it into a book.
Well, your company today is in that same workshop. AI makes producing text almost free, and the question becomes: of what you publish, what is truly yours?
What Italian law says
Law 132/2025, published in the Gazzetta Ufficiale on September 25, 2025 and in force since October 10, 2025, amends the copyright law (no. 633 of 1941) in its article 25. It does two things:
- in article 1 it states that protected works of the intellect are human ones, and that they are protected "even where created with the aid of artificial intelligence tools, provided they are the result of the author's intellectual work" (our translation);
- it adds article 70-septies, which allows reproductions and extractions of works one has lawful access to, for text and data mining to train AI systems, in line with articles 70-ter and 70-quater. That concerns those who train models; for what you can do on your own site, read robots.txt and AI training.
The law does not say how much intellectual work is needed. That is the open point, and judges will decide it case by case.
What the United States says
In January 2025 the US Copyright Office published Part 2 of its report on copyrightability. In short:
- using AI to assist rather than stand in for human creativity does not prevent protection;
- copyright protects the original expression of a human author, even if the work contains AI-generated material;
- it does not protect purely AI-generated material;
- with currently available technology, prompts alone do not give enough control over the output;
- the author can protect what is theirs and perceptible in the output, the creative selection, coordination and arrangement of materials, and creative modifications.
Courts point the same way. In Thaler v. Perlmutter the federal appeals court confirmed that a human author is required; the Supreme Court declined to review the case on March 2, 2026. Careful: declining to hear a case is not a ruling on the merits.
A practical case, as a typical scenario
Imagine an online shop with 8 people that has AI write 300 product pages. Nobody rereads them closely. A competitor copies them.
| What you did | How protected you are |
|---|---|
| One prompt, text pasted as is | Little: human input is hard to prove |
| AI draft, rewritten and corrected by a person | The rewritten part is yours |
| Selection and layout of images you chose | Creative selection may be protected |
| Photos, texts and logo you made, used as a base | Those are yours in any case |
This is a guide, not legal advice: doubtful cases should go to a lawyer. One thing is simple, though: your brand, name and product data don't depend on this question, and are protected by other tools.
What to do on Monday morning
- Sort the content that matters. List what you really need to protect (logo, site copy, manuals, photos) and for each write who made it and with which tool.
- Have a person go over everything you publish. Rewriting, choosing and correcting is the contribution the law looks for. Keep a simple trail: AI draft, final version, date and name of who worked on it.
- Check contracts with whoever produces content for you. Agencies and freelancers who use AI: ask in writing what they deliver and who owns the rights. For internal rules, see rules for employees using ChatGPT and, for the duty to flag synthetic content, labeling AI-made content.
Johannes de Spira was given five years of protection and used few of them, since he died in 1470; but the idea stayed: protect whoever puts in the work. The same principle holds today. If you want to work out how to organize all this in your company, book a free intro call. You can also see our services and the page for independent professionals.
The Professor of Penzai

