IntegrAuth Academy
Identity, security and practical AI — free, from zero. Byte-sized lessons, real diagrams, hands-on labs, zero fluff. Two programs, each with its own exam and certificate.
English is hard? Your browser can translate this page into Tamil, Telugu, Kannada, Malayalam or Hindi. Tech words stay in English so you learn the real names.
How?
- Chrome on phone: ⋮ menu → Translate
- Chrome on computer: right-click → Translate to …
- Edge: the translate icon in the address bar
Who owns AI output?
Priya's team wants to launch with an AI poster, slogan and code. Do they own any of it? Your goal: ask three questions before you publish.
🧪 Interactive lab — enable JavaScript to play with this one.
- "Who owns this?" is really three questions.
- In the US, prompts alone do not make you an author.
- Your own text, sketches, choices and edits can be protected.
- An output that copies someone's work is still theirs.
- Terms can only pass on rights that exist.
The law gives creators rights over their original work, like text, images or code. This is copyright. It protects how something is expressed, not the idea. This is not legal advice. Laws differ by country.
1. Can anyone own it?
In the US, only a human can be an author. This is human authorship. A prompt alone does not make you the author, even after dozens of tries. Effort is not authorship. What counts is what you create: your text, your sketch, and how you choose and arrange things. Write the story yourself, and your text and layout can be protected. Keep your drafts as a record.
2. Does it copy someone else?
A model can repeat parts of its training data almost word for word. This is memorization. Image tools can also copy famous characters. If you publish a copy, it is your problem, however it was made.
3. What did you agree to?
Tool terms often give you their rights in the output, "if any". A contract can't create a copyright the law does not give. Some paid plans promise to defend you and pay if you are sued over an output. This is an IP indemnity. Read your own plan.
Now help Priya make her poster her own.
🧪 Interactive lab — enable JavaScript to play with this one.
Go deeper: court cases
Stephen Thaler asked the US Copyright Office to register an image. He named his AI system as its only author. The Office said no, and courts agreed. In March 2026, the Supreme Court declined to hear his appeal. The appeals court added that works made with AI's help can still be protected. It did not say how much is enough.
In January 2025, the Office said prompts work like instructions. They pass on ideas, and today's tools don't give users enough control. In 2023, it protected the comic book Zarya of the Dawn. It protected the human-written text and the layout of words and pictures. It did not protect each image Midjourney made.
To register a work in the US, you must disclose AI material that is more than minimal.
Go deeper: other countries
In 2023, a court in Beijing, China, protected an image because the user chose the prompts and tuned the settings. The US says that is not enough.
The UK has a 1988 rule that protects "computer-generated" works with no human author. In March 2026, the UK government proposed removing it.
India's law has a similar rule. For a "computer-generated" work, the author is the person who causes it to be created. As of October 2026, no court has applied this rule to AI output. Check the rules where you publish.
Go deeper: training and fair use
A separate fight is about training models on books and images without asking. In the US, it depends on fair use. Courts weigh four factors, like harm to the original's sales. EU law allows text and data mining unless the rights holder opts out.
As of October 2026, one US appeals court has said training a non-generative AI was not fair use. No appeals court has ruled on generative AI, and lower courts disagree. In July 2026, a Delhi court refused a news agency, ANI, an early order against OpenAI. It saw the training as likely fair dealing. The full case goes on. These fights target the model makers, not you.
Go deeper: terms and indemnities
Many terms say you own the output "to the extent the law allows". Some warn that output "may not be unique". Another user can get something similar.
Terms can limit use, too. Some forbid using output to build a model that competes with the provider. See Open and closed models.
Microsoft, Google and OpenAI announced IP indemnities in late 2023 for paid business services. They usually leave out free plans. They come with conditions, like keeping content filters on and not knowingly using output that copies someone.
Go deeper: code licenses
Open-source code comes with a license. Permissive ones, like MIT, ask you to keep the copyright notice. Copyleft ones, like the GPL, ask more. When you share your software with others, you must share your changes under the same license. If an AI tool copies licensed code, those rules apply to you when you ship it.
Some coding assistants check suggestions against public code. GitHub Copilot, for example, can block matches or show their license. License scanners, like the open-source ScanCode Toolkit, list the licenses in a project.
At Priya's company, a scan flagged a function copied from a GPL project. The assistant rewrote it from a plain description.
"The terms say I own it" is not a copyright. And neither one means it is safe to publish.