The AI Act is the first EU regulation governing the development, supply and use of artificial intelligence. If you design, deploy or use AI, the Regulation imposes specific duties on you. Failure to comply may mean substantial financial penalties and administrative liability. This article introduces the AI Act, its implementation timeline and the definition of an AI system, which determines whether the rules apply to you at all.<\/p>\n
Before looking at the AI Act itself, it is worth considering the challenges that led to its creation. They help explain why the European Union decided to regulate the use of artificial intelligence.<\/p>\n
Key issues that shaped the rules include:<\/p>\n
AI is one of the few technologies whose output cannot always be traced and explained clearly. Opaque algorithms of this kind are called a “black box”.<\/p>\n
<\/p>\n
Businesses need to act on two levels:<\/p>\n
In other words:<\/p>\n
Otherwise, you could unknowingly disclose confidential data, for example to a server in China. That is an uncomfortable situation for any business.<\/p>\n
The black box issue is directly connected to another challenge: data<\/strong>, particularly processing it in accordance with the GDPR<\/a><\/strong>.<\/p>\n In simplified terms, AI learns from data provided by its developers and users. Many systems also use publicly available online content, making it harder to control exactly what they use and how they arrive at their answers.<\/p>\n <\/p>\n If your business plans to use AI, ask these questions before processing any data:<\/p>\n These are questions to resolve in advance. The answers affect whether your business complies with the GDPR<\/a><\/strong> and whether it faces real risks, including penalties, loss of client trust and data breaches.<\/p>\n For a business, two things are at stake:<\/p>\n Another important issue is copyright in AI-generated work. Who has rights to images, text, code or other content generated by AI? Can it be lawfully used, protected and even sold?<\/p>\n These questions matter to businesses that create content with AI and those planning to commercialize it.<\/p>\n <\/p>\n Two potential problems are particularly important:<\/p>\n 1. Risk of infringing someone else’s copyright.<\/strong> AI may, for example, produce an image very similar to an existing one, such as a stock photograph or another artist’s work. You may not even know it is a “plagiarism”, but if a client publishes it, the client may be liable for copyright infringement. AI does not create from nothing; it learns from other people’s work.<\/p>\n <\/p>\n 2. Uncertainty about rights in your own output.<\/strong> Copyright protects works created by people. If an image, text or code is created solely by AI, there is formally no author and therefore no copyright. If your employees create something “with the help of AI”, there is no simple answer either. It is necessary to assess how creative the person’s contribution is and whether the result is individual. Case law in this area is still developing. Hopefully, clear rules will emerge for determining when copyright arises in a work created by a person using AI tools.<\/p>\n In practice, your business should ensure:<\/p>\n At the outset, it is important to note that the AI Act does not, in principle, regulate the copyright issues described above.<\/p>\n Artificial intelligence can now operate fully autonomously. It may decide whether to hire or dismiss an employee, grant credit, assess financial risk or user behavior, profile clients, and even influence political views and purchasing decisions.<\/p>\n Its potential influence on people, businesses and entire societies is considerable and not always positive.<\/p>\n That is why it became necessary to clarify what AI may be used for, which applications require additional oversight and which are completely prohibited<\/a><\/strong>.<\/p>\n The AI Act is intended to provide that framework.<\/p>\n The fact that you did not create an AI system does not mean you have no responsibility for it. If your business uses AI, legal risk may concern you too. AI makes decisions, advises and recommends, and sometimes acts on a person’s behalf. If an error, harm or complaint occurs, someone must take responsibility.<\/p>\n <\/p>\n The role your business plays matters:<\/p>\n The AI Act therefore specifies duties for AI providers and deployers. Their scope depends on the risk level of the system and on whether it qualifies as an AI system under the Regulation.<\/p>\n The Artificial Intelligence Act entered into force on 1 August 2024. Not all its rules applied immediately; implementation was phased over time to allow businesses and institutions to adapt without excessive operational disruption.<\/p>\n <\/p>\n The timeline is as follows:<\/p>\n If you create or use software that does not meet the definition of an AI system, the AI Act does not apply to you. In that situation, the new rules do not apply and you do not need to make changes.<\/p>\n The AI Act defines an artificial intelligence system as:<\/p>\n “a machine-based system designed to operate with varying levels of autonomy, that may exhibit adaptiveness after deployment and that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations or decisions that can influence physical or virtual environments.”<\/em><\/p><\/blockquote>\n Check whether your software meets these conditions:<\/p>\n If your system meets these conditions, even partially, it is worth assuming that it falls within the scope of the AI Act and that it requires at least a preliminary compliance analysis.<\/p>\n\n
\n
Copyright and related rights<\/h3>\n
\n
What is permitted and what is prohibited in AI use?<\/h3>\n
Responsibility for AI<\/h2>\n
\n
When does the AI Act apply?<\/h2>\n
\n
\n
\n
Definition of an artificial intelligence system under the AI Act<\/h2>\n
When does a solution meet the definition of an AI system?<\/h3>\n
\n