copyright
Copyright is the rule system under which the creator or rights holder decides how a text, image, music, or program is used and can charge for it.
In plain terms
Copyright is like the key to your home: whoever holds the key decides who gets in. If you write a poem, photograph a landscape or write a program, you decide who may copy it, publish it or adapt it, and for how much.
Protection is automatic; there is no need to register or pay anywhere. However, the idea itself is not protected, only the expressed work: the concept of a "wizard school novel" is free, but the text of a specific novel is not.
Artificial intelligence has stirred up this field at two points. One question is whether it is permissible to train language models and image generators on millions of other people's works without permission and payment. The other is who owns what the machine creates, and whether anything protects it at all.
How it works
- A work becomes protected the moment it is created if it has an individual, original character. No registration or © symbol is required.
- The author is entitled to two types of rights: moral rights (for example, to have their name displayed) and economic rights (permission to copy, distribute or adapt, typically for a fee).
- Anyone wishing to use the work requires permission, i.e., a license. Exceptions exist for cases specified by law, such as quotation, private copying, or certain educational uses.
- In the EU, the text and data mining exception applies to AI training. Lawfully accessible content may be analyzed by machines. For commercial analysis, the rights holder can prohibit this (opt-out), and if the content is available online, the prohibition must be machine-readable. No such prohibition is possible for scientific research.
- Protection in the European Union lasts for 70 years from the end of the year following the author's death. After that, the work enters the public domain and anyone can use it freely.
An everyday example
You create a poster for your company's summer event using an image generator, and the result strikingly resembles a well-known cartoon character. Even though the machine 'drew' it, if you publish it, the rights holder of the character can demand its removal. Meanwhile, it is not certain that your own poster could be protected if someone copies it.
Who is affected
A common misconception
Limits and risks
- Copyright varies by country: the American 'fair use' is a flexible standard, while in Europe there is a closed, statutory list of exceptions, so an American ruling is not binding here.
- Many questions are still before the courts at the end of 2026, so an answer given today on the legality of training may change in a year or two.
- AI service providers' terms of use do not always accept liability for the output, so in case of infringement, the publisher, that is you, may also be held responsible.
- Copyright does not protect ideas, style, facts, or data, so imitating a creator's style alone does not offer reliable legal protection.
Similar, but not the same
- szabadalom – A patent protects a technical invention for a shorter period, after filing and examination, while copyright protects works automatically.
- trademark – A trademark links a brand name or logo to a company so buyers do not confuse products, and it does not protect the work itself.
- data protection – Data protection concerns your personal data, while copyright concerns your creations; a photo of you involves both, but each protects a different person.
Related concepts
text and data mining, licenc, public domain, fair use
AI regulation, training data, generative AI, EU AI-rendelet, open source, watermarking
A brief history
The first modern copyright law was the English Statute of Anne, which came into force in 1710, and the international foundations were laid by the 1886 Berne Convention. In Hungary, the first copyright law was Act XVI of 1884; today, the area is regulated by Act LXXVI of 1999 (Szjt.). The EU's 2019 directive harmonized the text and data mining exception at the EU level, which has become one of the main legal frameworks for AI training.
Where things stand
Copyright itself has long been valid everywhere, but its application to AI is still evolving by the end of 2026: lawsuits over training are ongoing in several countries, while publishers and AI companies are concluding more licensing agreements. The EU AI Act requires providers of general-purpose models to publish a copyright compliance policy and a summary of the training data.
In the news
- Eminem's publisher and the Elvis estate both sue AI music platform SunoNov 7, 2026 · Jog
- US federal judge dismisses music publishers' direct copyright claim against Dario AmodeiNov 5, 2026 · Jog
- Anthropic copyright settlement could benefit Herald Press authorsNov 5, 2026 · Jog
- LG AI Research tool checks copyright risk in training data 45 times fasterNov 4, 2026 · Jog
- EU Court of Justice adviser says Google's Gemini may infringe copyrightNov 3, 2026 · Agent Age
- EU court adviser says Google's AI chatbot may infringe copyright in press contentNov 3, 2026 · Marketing
- Ross Intelligence seeks new hearing in copyright case against Thomson ReutersNov 3, 2026 · Jog
- Hollywood, creators' groups and publishers back New York Times in OpenAI caseNov 2, 2026 · Jog
Also known as
copyright · copyright · copyright · copyright protection · © · Szjt. · author's rights · copyright law
Sources
This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.