AI Act – What is the Artificial Intelligence Act and who does it apply to? Key information.
The European Union has prepared the Artificial Intelligence Act, known as the AI Act, which aims to ensure the development and use of artificial intelligence in a safe, reliable, and transparent manner. What does it regulate, and who will it cover? This article contains the key information.
As of the time of writing this article (early February 2024), there is still no precise information on when the European Parliament will officially approve the AI Act. Regardless of when that happens, one thing is certain – Poland should already be working on new regulations, and businesses should be preparing for their implementation. Especially since the agreement on the new AI-related rules has already been reached, and we know who and what they will affect.
The European Union's Artificial Intelligence Act, or what the AI Act is
The European Union’s Artificial Intelligence Act, proposed in April 2021 by the European Commission, is the first and most comprehensive legal framework aimed at regulating the use of artificial intelligence systems. Why is this important? Because their application is associated with various risks that have already been identified, and the list of these risks will certainly expand in the future as AI tools continue to develop.
What does the AI Act regulate?
The new regulation classifies artificial intelligence systems based on the associated risk. Although many of them pose minimal risk, they must be properly assessed. To this end, a risk-based approach has been applied, leading to horizontal regulations, which are applicable across all sectors.
Depending on the risk classification, three solutions are adopted. Specifically, the system may be banned; additional requirements may be imposed, or users interacting with the AI must be notified about the interaction. It is worth noting that the EU allows certain exceptions for high-risk systems, such as permitting the system for law enforcement purposes.
Who does the AI Act apply to?
The law has significant consequences for organizations that develop or use artificial intelligence systems. Moreover, it applies regardless of their location — what matters is that the tool is available/used within the EU.
The AI Act applies to any system that is:
- sold or offered on the EU market;
- put into use or utilized within the borders of the EU.
It is important to note that there is no size limitation for businesses. The law applies to both global tech giants and local developers selling AI software. All ventures related to the use and offering of artificial intelligence must meet the same standards: safety, reliability, and transparency.
The EU strategy aims to ensure the credibility of artificial intelligence and transparency at every stage of its development. This, in turn, will make it easier to comply with the regulations and help align all parties involved in AI development. It will also facilitate investments and innovation in this field.
What are the benefits of adopting the AI Act?
Every action aimed at securing the interests and privacy of individuals is invaluable, especially in light of concerning reports of AI being used for cyberattacks.
The EU legislation aims to protect against all negative consequences of artificial intelligence. Therefore, it somewhat forces organizations using such systems to ensure that they do not discriminate against people, mislead them, or that their applications are not used in an unethical or even illegal manner. Users should be aware that they are interacting with AI, regardless of the type of system—whether it’s an application generating images, audio or video content, or manipulations through deep fake technology.
Challenges faced by entrepreneurs
Entrepreneurs may feel concerned about the need to implement the requirements of the AI Act. This is understandable, as the consequences of this legislation can be drastic and costly. Especially for companies that develop and supply software classified as high-risk – they will still be able to continue their development, provided specific requirements are met. This mainly applies to sensitive areas such as education, human resources, critical infrastructure, and law enforcement.
How compliance with EU regulations is checked:
- An audit to check the AI system’s compliance with the EU law (the assessment must be repeated in case of significant modifications to the system).
- Checking how risk is monitored, including technical documentation, records, human oversight, and management.
- Assessing the system’s impact on fundamental rights.
- Evaluating the transparency of the results and the model.
- Assessing security and the accuracy and robustness of the models used.
Are there penalties for failing to meet the requirements? Yes. Depending on the violation and the size of the company, fines will range from €35 million or 7% of global turnover. Failure to provide correct information will result in a fine of up to €7.5 million or 1.5% of the organization’s global annual revenue.
How to prepare for the AI Act?
The AI Act is still in the process of being finalized, and its final version and implementation date are not yet known. However, it is worth starting preparations now. This will ensure that the organization is fully ready, and the implementation process will be smooth, without disrupting operations. Moreover, it will be completed on time, helping to avoid financial penalties.
It is best to entrust this task to an experienced consulting firm specializing in such implementations. An example of such a company is PBSG. We will conduct the necessary compliance assessments and risk analysis; we will also check the system for cybersecurity.