The AI Act: An overview

The EU Artificial Intelligence Act (AI Act) is a recently implemented regulation aimed at improving the functioning of the internal market and promoting the uptake of human-centric and trustworthy AI systems (AI Act, Art.1). As it’s the first of its kind, the Act is poised to be a transformative piece of legislation, setting a precedent for how AI should be regulated.

The AI Act clarifies the concept of AI in Art. 3(1) by defining it as “a machine-based system that is designed to operate with varying levels of autonomy and 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.”

The regulation provides detailed guidance on several key aspects (Art. 1(3)):

• harmonised rules for placing AI systems on the market, putting them into service, and the using them in the Union
• prohibitions of certain AI practices
• specific requirements for high-risk AI systems and obligations for operators of the systems
• harmonised transparency rules for certain AI systems
• harmonised rules for placing general-purpose AI models on the market
• rules on market monitoring, market surveillance, governance and enforcement
• measures to support innovation, with a particular focus on SMEs, including startups

The Act classifies AI into four categories, two of which are highly relevant: Prohibited AI and High-risk AI.

Prohibited AI (Chapter II, Art. 5)

In short, the Act deems AI systems as prohibited if they:

• use subliminal, manipulative, or deceptive techniques to distort the behaviour of an individual or group by disabling them to make an informed decision;
• exploit the vulnerability of individuals or groups with the aim of distorting their behaviour to cause them significant harm;
• evaluate or classify individuals based on their behaviour or personal traits leading to unfavourable treatment of those individuals;
• form risk assessments of individuals based on their personality traits and characteristics to predict the risk of that individual of committing a criminal offence. This prohibition doesn’t apply in cases with objective and verifiable facts directly linked to criminal activity;
• develop facial recognition databases using facial images from the internet or CCTV footage;
• infer emotion from individuals in a workplace or education, except for medical or safety reasons;
• use biometric identification to deduce or infer their race, political opinions etc; or
• use real-time remote biometric identification systems in public spaces for the purposes of law enforcement, except in cases of victims of abduction, prevention of threat to life or physical safety, or the localisation or identification of an individual suspected of having committed a criminal offence.

For example, a prohibited AI system under this Act would be a system using subliminal messaging to influence individuals’ purchasing decisions.

High-risk AI (Chapter III, Art. 6)

According to the Act, AI systems are high-risk if they:

• are a product or are intended to be used as a safety component of a product; or
• are a product that is required to undergo a conformity assessment, with the intention to be placed on the market.

A high-risk AI system under this Act would be a system that controls the operation of industrial plants or automated transport systems, for example.

Scope of application

In terms of its application, the name of this Act may be slightly misleading. The Act can in fact apply to AI outside the EU.  

Art. 2(1) stipulates that the regulation applies to AI systems in the union and in third countries. So, regardless of where in the world an AI system is based, the regulations will only cease to apply if the output or use of that system falls outside the EU. All other AI systems based and used in an EU member state will apply. So, for instance, if an AI system is based in the US but its output is in the Czech Republic, it will have to follow the regulation.

Governance

As will be the case with the other member states, the Czech Republic will have various forms of governance within the regulation, ranging from having representatives of each state on the AI Board to having a designated national competent authority.

The aim of the representatives is to provide assistance and advice to the Commission, ensuring the Act is applied effectively.

Current AI ecosystem in the Czech Republic

Over the past few years, the Czech Republic has seen an increase in AI development, showcased by various state initiatives and strategies led by institutions liked the Czech AI Association (Česká Asociace Umělé Inteligence) and the Czech Ministry of Industry and Trade.

The Czech Ministry of Industry and Trade has recently approved the National Artificial Intelligence Strategy of the Czech Republic 2030 (NAIS) with the aim of advancing the country's economy and society by continuously employing and developing AI. This document acts an update of the original AI strategy of the Czech Republic (2019), while also taking into consideration the recent EU AI Act discussed above.

Though this regulation isn’t legally binding, it is a key player in the Czech Republic’s AI sphere.

The strategy sets out seven primary goals for the country to meet by 2030:

• research, development and innovation
• education and professional training
• skills and AI impacts on the labour market
• ethical and legal aspects
• security aspects
• industry and business
• public administration and public services

The AI Act and the NAIS: How do they compare and how will they shape the Czech AI landscape?

Some of the recurring factors we see in these two documents are innovation of AI, ethical and legal aspects, industry and business. The AI Act will help the Czech Republic reach its strategy goals simply by being implemented into national law.

It can be expected that according to the classification system provided in the Act, certain AI systems will be restricted in some way. In the case of prohibited AI systems, the act will apply six months from its entry into force (2 February 2025). After this period, if companies don’t comply with the prohibition of these AI practices, they could be fined up to EUR35 million or up to 7% of their total worldwide annual turnover (Art. 99(3)). This means that, as of now, businesses in the Czech Republic and other EU member states have less than six months to ensure they comply with the conditions set out in this regulation.

It is reasonable to conclude that the introduction of this Act will benefit the Czech Republic rather than cause detriment. Given the complexity and scope of AI regulation, smaller countries like the Czech Republic would face significant challenges in developing such comprehensive legislation independently. So it is advantageous that a larger institutional body, such as the EU, has assumed this responsibility, alleviating the regulatory burden on member states.

What to expect in the upcoming months

2024

• 2 November 2024: Among other member states, the Czech Republic will have to identify the authorities and bodies responsible for fundamental rights protection (Art. 77(2)).

2025

• 2 February: Prohibitions on certain AI systems start to apply (Art.113(a)).
• 2 August: Rules on notified bodies, GPAI models, Governance, Confidentiality, and Penalties start to apply (Art.113(b)).
• 2 August: Deadline for member states to designate national competent authorities. The Czech Republic is expected to designate this role to one of the following bodies: the Ministry of Industry and Trade, the Office for Personal Data Protection, the Czech Governmental Office, or the Czech Telecommunication Office.

For a more detailed timeline see Implementation Timeline | EU Artificial Intelligence Act