Právní ochrana AI-generovaného obsahu

Basic Question: Can Artificial Intelligence Be an Author Under Czech Legislation?

The first pillar is the definition of an author. Under Czech legislation, only a natural person—a human being who created the work—can be an author. Artificial intelligence, no matter how advanced, is software, not a human. For this reason, it is automatically excluded from the role of author.

The second pillar is the definition of the work itself. A copyrighted work is only the unique result of the creative activity of the author. Case law has long interpreted this concept as an activity that bears the imprint of the author's personality and has an unmistakable, distinctive character. It is a mental process unique to humans.

The machine process of generating content, which is based on data analysis and algorithms, does not meet this condition. As a result, content created purely by artificial intelligence without significant human creative intervention is not a copyrighted work under current Czech legislation. This fundamental mismatch between legal philosophy and technological reality is the source of all further complications.

When AI Creates, Who Holds the Rights? The User, the Developer, or No One?

So, if AI cannot be an author, who owns the rights to the content it generates? Many users mistakenly believe that authorship automatically transfers to them, especially if they enter a creative and detailed instruction, a so-called "prompt", into the AI. However, the legal reality is more complex and often dangerous for businesses.

Why a Creative "Prompt" Is Not Enough?

A recent decision by the Municipal Court in Prague shed light on this issue. The court stated that a user prompt, however original it may be, is from a legal perspective merely an idea, concept, or procedure. And the Copyright Act explicitly excludes ideas and concepts themselves from protection. Creative activity is the actual execution, which in this case is carried out by the AI, not a human.

The Danger of "Public Domain"

If the AI output is not a copyrighted work because it has no human author, it becomes a so-called public domain work. This means it is not protected by copyright and anyone can freely copy, distribute, and commercially exploit it without any penalty.

This poses a huge risk for your company. A logo, marketing text, or unique graphics that you have generated and invest in establishing can be copied and used by your competitors with impunity. You have no legal tool to defend yourself.

How to Obtain Copyright?

The only way to obtain copyright protection for content created with the help of AI is through significant and demonstrable human creative intervention in the already generated output. Minor proofreading is not enough. It must be a modification that in itself meets the characteristics of creative activity. Copyright then applies only to these human modifications, not to the original, machine-generated base.

Many companies rely on the terms and conditions of AI tools, which often state that "ownership" of the content belongs to the user. However, this is a dangerous misconception. This is merely a contractual agreement between you and the provider, meaning that they themselves will not claim the content. But this private commitment cannot establish statutory copyright protection that would be enforceable against third parties.

Hidden Risks You Must Know: From Plagiarism to Data Leaks

In addition to uncertainty regarding ownership of rights, the use of AI in companies brings a number of other, very specific risks. Ignoring them can be costly, not only financially but also in terms of reputational damage.

Input Risk: The Training Data Problem

AI models learn from huge amounts of data, which are often obtained by automatic downloading from the internet (so-called web scraping). This process involves processing millions of texts, images, and other copyrighted works, usually without the consent of their authors.

Output Risk: Unintentional Plagiarism

Because AI learns from existing works, there is a risk that it will generate content that is too similar to an already existing protected work. Your company could thus unknowingly commit copyright infringement and face a lawsuit, even if it acted in good faith.

Liability for Errors and Misinformation

AI models are prone to so-called "hallucinations"—generating factually incorrect, misleading, or even defamatory information. Legal liability for damage caused by publishing such content always rests with your company, not the AI provider. From a legal standpoint, AI is merely a tool, and the person who uses and publishes its outputs is fully responsible for them.

Leak of Sensitive Data and GDPR Violations

Any information you enter into publicly available AI tools—whether internal strategies, client data, or trade secrets—is transferred to a third party, i.e., the operator of the service. This can violate not only non-disclosure agreements (NDAs) but also the strict rules of the GDPR regulation if personal data is involved. This is particularly risky with operators based in the US, where a different data protection regime applies.

The AI Act Is Coming: What New Obligations Await Your Company?

The debate on AI is shifting from copyright uncertainties to new, direct, and very specific regulation. The EU Artificial Intelligence Act, known as the AI Act, is the first comprehensive legislation of its kind in the world and brings a number of new obligations for companies. While copyright is passive (addressing whether something is a work), the AI Act is proactive—imposing obligations on companies simply for using AI.

The AI Act introduces a risk-based approach, classifying most generative tools, such as ChatGPT, into the limited risk category, which is subject to specific transparency obligations.

Key obligations for your company include:

• Labeling AI content: You must clearly and comprehensibly label all content (texts, images, videos, audio) that has been created or substantially modified by artificial intelligence. The aim is to prevent public deception and ensure that users always know they are interacting with artificial content.

• Publishing training data: Providers of generative models will have to publish summaries of copyrighted data on which their model was trained.

• Preventing illegal content: Providers must design their models to actively prevent the generation of illegal content.

Violations of these rules carry astronomical penalties, which can reach up to EUR 35 million or 7% of the company's global annual turnover, whichever is higher.

How to Protect Yourself? Practical Steps for Safe Use of AI in Your Company

The "do nothing" strategy is unsustainable and dangerous in the current environment. A proactive approach to managing AI-related risks is not only a necessity but also a competitive advantage. Here are the key steps that every responsible company should take.

Introduce an Internal Policy (AI Policy)

This is an absolutely crucial and necessary step. A clearly defined and written internal policy for using AI is your main defense mechanism. Without it, it is difficult to prove that you have fulfilled your duty of prevention and that an employee violated the established rules. The policy should at least include:

• A list of approved and prohibited AI tools.

• A precise list of data that must not be entered into AI (trade secrets, personal data, client information).

• An obligation to verify the factual accuracy and originality of outputs.

• Rules for labeling AI-generated content in compliance with the AI Act.

Train Your Employees

Most security incidents associated with AI do not arise from technology failure, but from human error due to ignorance. Regular and demonstrable training of employees on risks, internal policy rules, and obligations under the AI Act is a key investment in your company's security.

Read License Terms Carefully

The conditions for commercial use differ dramatically for individual tools.

• OpenAI (ChatGPT, DALL-E): Generally allows commercial use and contractually transfers rights to the user.

• Midjourney: Conditions commercial use on more expensive plans (Pro or Mega) for companies with an annual turnover of over USD 1 million.

• Microsoft Bing Chat Enterprise: May have different, often more restrictive conditions that need to be thoroughly reviewed.

Conduct an Audit of Used Tools

Map out which AI tools are actually used in your company and for what purpose. Many employees may use various applications without management's knowledge (so-called shadow IT), which represents an uncontrolled risk. An audit will help you gain an overview and set rules.

International Reach: Solutions for Global Business

For companies operating in international markets, the situation is even more complex. Legal approaches to AI regulation differ globally, with the two main models represented by the EU and the US.

The European Union has chosen the path of comprehensive, preventive top-down regulation (the AI Act), which places great emphasis on the protection of fundamental rights and transparency. Conversely, the United States applies a more sector-specific, market-oriented approach without unified federal legislation. They rely on existing laws and the decision-making practice of authorities such as the U.S. Copyright Office.

Even in the US, however, there is a strict requirement for "human authorship" as a condition for granting copyright protection. The U.S. Copyright Office refuses to register works created purely by machine and individually assesses the degree of human "creative control" over the resulting work.

This legal fragmentation means that a strategy that is legal and safe in one jurisdiction may be risky in another. Thanks to the ARROWS International network, built over ten years, we address issues with an international element practically every day and can ensure that your global AI strategy complies with various legal systems.

Conclusion

Artificial intelligence is undoubtedly a revolutionary tool with huge potential. However, its implementation brings complex legal challenges that cannot be ignored—from copyright to regulatory obligations and contractual risks. A proactive legal approach is not a cost, but an investment in a safe and sustainable future for your company.

At ARROWS, we address these issues on a daily basis. Our experience in long-term service provision to more than 150 joint-stock companies, 250 LLCs, and 51 municipalities and regions allows us to understand your needs in the context of real business. We pride ourselves on speed, high quality, and practical solutions.

We offer comprehensive legal services in the field of AI, which include the preparation of internal policies and contracts, professional training, representation before courts and administrative authorities, as well as ensuring compliance with the AI Act and GDPR. In addition, we are happy to connect our clients if we see interesting business or investment opportunities.

Do not tackle these complex issues alone. Put yourself in the hands of experts who will provide you with legal certainty and allow you to innovate with confidence. Contact us for a non-binding consultation and a tailor-made solution for your company.