copyright infringement

basicwidespread

Copyright infringement occurs when someone copies, distributes, adapts, publishes online, or publicly performs another's protected work (text, image, music, code) without permission or a legal exception.

In plain terms

If you write a poem, photograph a sunset, or compose a song, the work is protected from the moment of its creation, and you do not need to register it anywhere. You decide who may copy it, publish it, or adapt it. Copyright infringement occurs when someone bypasses this decision.

It is like someone printing your photo on T-shirts and selling them at the market. You keep your image, but others profit from it, even though you should be the one deciding. A key distinction is that copyright protects the expression, not the idea: anyone can write a novel based on the 'wizard school boy' concept, but no one may copy the Harry Potter text.

Artificial intelligence has brought this back into the news at two points. One is the input: was it lawful to train models on millions of books, articles, and images without the creators' permission? The other is the output: what if the machine produces something that resembles an existing work too closely?

Not all unauthorized use is illegal. The law recognizes exceptions, such as quoting with source attribution, private copying in certain cases, or educational use. In Hungary, this is called free use, while in the United States, the more flexible fair use principle serves a similar role.

How it works

First, you must decide whether the work is protected: is it an individual, original creation, and does protection still apply, which in the EU generally lasts for 70 years after the author's death.
Then check whether there was any use that required permission: copying, distribution, adaptation, online publication or public performance.
If such a case exists, they examine whether there is permission (license, contract) or a legal exception, such as quotation or a private copy.
Training is a separate issue in AI. In the EU, text and data mining of lawfully accessible works is generally permitted. Rights holders can opt out via a machine-readable notice on works made available online, which developers must respect. However, they cannot prohibit text and data mining for scientific purposes by research institutions and cultural heritage organizations.
If infringement is established, the rights holder can demand that it stop, that the infringing content be removed, that profits gained through the infringement be returned, and compensation. If the infringement causes financial loss, it may also constitute a misdemeanor or a crime. Penalties are more severe for commercial-scale infringement.

An everyday example

You find a good photo online and post it on your company's Facebook page because 'it was already on the internet.' A few months later, you receive a payment demand from the photographer's lawyer. The same can happen if you have an AI image generator create an advertisement image that looks exactly like a well-known cartoon character.

Who is affected

marketerFor images, music, and AI-generated materials used in campaigns, the primary responsibility lies with the publisher, and this responsibility cannot be automatically shifted to the tool manufacturer. → Marketing

parent, student, teacherWhen incorporating text and images into essays, presentations, or teaching materials, you must know what falls under citation and the educational exception. → Education & Family

developerA coding agent or other AI tool may suggest a code snippet from an open-source project whose license terms (such as attributing the author or making the source code available) you must comply with. If this is omitted, the infringement enters your product. → Code

entrepreneur, executiveIf content is created with AI in your company, the legal risk remains with the company, so it is advisable to establish a rule on what must be checked before release.

A common misconception

MythWhatever AI generates is certainly original, so it can be used freely.

In realityA model can output text, images or code that is essentially identical to a protected work, and in that case the publisher may also be liable. Moreover, purely machine-generated content does not itself receive copyright protection in many countries.

Limits and risks

Rules vary by country: what fits under the US 'fair use' doctrine may still be illegal under Hungarian and EU law.
The legal assessment of AI training is not settled: courts have issued differing rulings, so do not take for granted that a tool is "legally compliant".
Naming the source is not permission in itself: writing the author's name does not stop copying the entire work from still being an infringement.
Some AI service providers promise legal protection to subscribers, but this is usually conditional and mostly does not apply to free plans.

Similar, but not the same

plagiarism – Plagiarism is an ethical violation: you present someone else's work as your own, which is plagiarism even if the work is no longer protected. Copyright infringement is a legal issue and can be committed even by citing the author's name. The two often go together.
trademark infringement – Trademarks protect brand names and logos from unauthorized commercial use, while copyright protects the work itself from copying.
patent hoarding – A patent protects a technical invention, and only if it has been filed and granted by the office. Copyright, however, arises automatically and protects the form of expression.

Related concepts

Broader

Narrower

piracy, unauthorized adaptation, infringing use of training data

On the edge

fair use, fair use, text and data mining, training data, plagiarism

A brief history

The first modern copyright law was the British Statute of Anne, adopted in 1710; international protection was established by the 1886 Berne Convention. Hungary's first copyright law was Act XVI of 1884; today, the area is regulated by Act LXXVI of 1999. Due to generative AI, a wave of lawsuits began from late 2022, mainly from 2023: authors, publishers, newspapers, music rights organizations, and image agencies successively sued model developers.

Where things stand

The concept itself is centuries old and well-developed, but its application to AI is still taking shape at the end of 2026. Some American and European lawsuits have concluded with verdicts or settlements, but no unified direction has emerged. Since August 2025, the EU AI Act has required developers of general-purpose models to have a copyright policy, respect rights holders' opt-outs, and publish a summary of their training data.

Status: 2026 November

In the news

Appears in 13 Agent Age stories; the curve shows weekly occurrence in the news magazine.

Elvis Presley rights holders sue Anthropic over 274 songs used to train AIOct 26, 2026 · Jog
Owners of Elvis Presley catalogue sue Anthropic over 274 songsOct 24, 2026 · Jog
Canadian news publishers sue Anthropic for copyright infringement in OntarioOct 19, 2026 · Jog
Canadian news publishers sue Anthropic over alleged AI copyright infringementOct 14, 2026 · Jog
USA Today sues OpenAI for more than $250 million over ChatGPT copyright infringementOct 12, 2026 · Jog
USA Today publisher sues OpenAI for $250 million over copyright infringementOct 8, 2026 · Jog
ANI v. OpenAI: Indian analysis of AI training and fair dealingOct 6, 2026 · Jog
26 local publishers sue OpenAI and Microsoft over copyright infringementSep 21, 2026 · Jog

Also known as

copyright infringement · copyright infringement · copyright violation · piracy · copyright infringement · copyright violation · piracy

Sources

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.