Artificial intelligence gains a new legal framework! On 13 March 2024, the European Parliament adopted groundbreaking regulation on AI. Here are the key points regarding the world’s first comprehensive legal AI act regulating the development and use of AI systems.
AI Act – Significance of the Regulation
The regulation will significantly impact the development and use of AI in Europe and around the world. The AI Act is a crucial step towards ensuring that artificial intelligence is used in a manner that is:
- responsible
- safe
- ethical
- and in compliance with human rights.
AI Act – the Subjective Scope of the Regulation
The AI Act covers a wide range of entities involved in AI systems, including:
- providers introducing AI systems to the market or making them available,
- users of AI systems,
- importers and distributors of AI systems,
- authorised representatives of non-EU providers,
- manufacturers introducing AI systems under their own brand.
The AI Act applies to both EU and non-EU entities.
AI Act – the most important aspects of use
The Regulation applies to AI systems that operate on the EU market, regardless of the entity’s location. The most important aspects of the AI Act include:
Users’ protection | The AI Act focuses on protecting the rights and safety of individuals in the EU affected by AI systems. |
Adapting to new obligations | Entities covered by the AI Act will have to comply with some new obligations. For example: risk evaluation of AI systems |
Classification of risks of AI systems | The regulation categorises AI systems into four risk levels: – unacceptable, – high, – limited, – minimal. Systems that pose an unacceptable risk will be prohibited. This concerns those based on subliminal techniques. Another example is using biometrics to identify people in a crowd. |
Requirements for high-risk schemes | High-risk schemes, such as those used in the healthcare, transport and justice sectors, have to meet a number of strict requirements. |
Innovation support | The European Commission is tasked with developing guidelines and standards for AI systems. It will also promote research and development in this field. |
AI Act – implementation rules
The AI Act introduces uniform implementation rules across the European Union. As a result:
- There is no need for additional national regulations. The EU regulation will be directly applicable in all member states.
- Consistent and transparent AI application principles throughout the EU for businesses and citizens.
AI Act will be directly applicable in all member states. This solution will be similar to the General Data Protection Regulation (GDPR).
AI Act – risk categories for the use of artificial intelligence systems
The AI Act establishes four risk categories for AI systems. It determines the level of threat posed by such systems to the rights and safety of individuals. Each category brings specific responsibilities and requirements for providers and users.
Risk category | Description |
Low risk | It includes AI systems that do not pose a significant risk to the rights and safety of individuals. For example: calculators, simple games, and recommendation systems based on popular products. |
Medium risk | It covers AI systems that may have a limited impact on the rights and safety of individuals. For example: chatbots, facial recognition systems, content filtering systems on social media. |
High risk | It includes AI systems that can have a significant impact on the rights and safety of individuals. For example: decision support systems for recruitment, credit assessment, adjudication in court cases. |
Unacceptable risk | It includes AI systems that pose unacceptable risks to the safety and rights of individuals. For example: social scoring systems and real-time biometric remote identification systems. These systems are banned. |
AI Act – high-risk categories
Annex III to the AI Act defines high-risk AI systems. These are the ones that may have a significant impact on the rights and freedoms of individuals. They include:
Biometric identification systems | E.g. facial recognition, fingerprints, iris scans, and other biometric features. Real-time remote biometric identificationsystems are prohibited |
Critical infrastructure management systems | Includes air traffic control systems, power grids, water supply systems, etc. Strict safety measures and human supervision are required |
Educational systems | Recruitment systems, student assessment and personalisation of learning. Transparency, fairness and non-discrimination are required |
Recruitment and employee management systems | Includes candidate assessment, staff selection and performance management systems. Transparency, fairness and non-discrimination are required |
Credit scoring and access to services | Includes credit risk assessment systems and access to other financial and public services. Transparency, fairness and non-discrimination are required. The right to explain the decision is also necessary. |
Law enforcement systems | Includes systems for crime prediction, evidence analysis, and arrest decision-making. Rigorous human oversight, transparency, and bias prevention are required. |
Migration, asylum and border control management systems | Includes systems for assessing asylum applications and granting visas. Rigorous human oversight, transparency, fairness, and non-discrimination are required. |
Justice systems and democratic processes | Includes systems for court rulings, predicting verdicts, and assessing recidivism risk. Rigorous human oversight, transparency, fairness, and non-discrimination are required. |
AI Act – obligations of entities using high-risk AI systems
The AI Regulation imposes specific obligations on entities using high-risk AI systems. These include:
- Evaluating risk and impact on fundamental rights;
- Implementing appropriate security and risk management measures,
- Ensuring transparency of the AI systems operations,
- Respecting individuals’ rights to obtain, access or correct information. Additionally, allowing deletion of data, and respecting processing restrictions.
- Enabling individuals whose rights have been violated to file complaints with supervisory authorities. This includes pursuing their rights in court
AI Act – advantages of use
The introduction of the AI Act will bring many benefits:
Legal confidence and a sense of security for citizens | Businesses and citizens will feel confident and safe about the rules on AI in all Member States. |
Cost reduction | Direct application of the rules will help to eliminate the need to comply with different national rules. |
Supporting innovation | Facilitating the development and use of AI technologies across the EU. |
AI Act – Timeline for Implementation
The AI Act will be implemented gradually over the next few years. Here is a detailed schedule:
Deadline | Scope of the provisions |
20 days after publication in the Official Journal of the EU (around 21 June 2024) | – entry into force of the regulation, – prohibition of the use of AI systems with unacceptable risks. For example: systems based on social scoring and remote biometric identification in real time. |
6 months from the entry into force (December 2024) | – the establishment of AI supervisory authorities in the Member States, – the launch of the European Group on Artificial Intelligence. |
12 months from entry into force (June 2025) | – transparency obligations for providers of general-purpose AI systems, – obligations on systems for monitoring and recording human intervention in the operation of high-risk AI systems, – certain provisions on penalties for violation of the Act. |
36 months from entry into force (June 2027) | – the full entry into force of the obligations for high-risk AI systems, including: – requirements for risk assessment and implementation of security measures, – obligations regarding human supervision and documentation, – rules for AI systems used in the system of justice and democratic processes. |
Please note that:
- these deadlines may still change,
- the exact dates of entry into force of the individual provisions will be announced later,
- The European Commission will develop guidelines and tools to facilitate the implementation of the Act.
AI Act – summary
The AI Act is a comprehensive regulation that aims to ensure the safety and ethics of artificial intelligence. It introduces several new requirements for AI systems. Yet, it also prohibits certain harmful practices. The AI Act will significantly impact the development and use of AI. Not only in Europe but also globally.
AI is a new and interesting topic. If you want to learn more about the AI Act, our experts are ready to help. Contact us!
FAQ – Questions and Answers about the AI Act
What is the AI Act?
It is the world’s first comprehensive legal regulation on the development and use of AI systems. The goal of the Act is to ensure that artificial intelligence is safe and ethical. Another aim is to ensure that AI complies with human rights.
What are the risk categories of AI systems?
The regulation divides AI systems into four risk categories:
– Unacceptable. These aresystems that pose unacceptable risks. For example, by being based on subliminal techniques or using biometric data to identify people in a crowd. These systems are entirely prohibited.
– High-risk. These are used in the healthcare, transport and justice sectors. They must meet a number of strict requirements.
– Limited. These are low-risk systems, e.g. chatbots and recommendation systems. They are subject to less strict obligations.
– Minimal. Minimal-risk systems, such as calculators and games. They are not subject to specific requirements.
What are the obligations for high-risk systems?
Users of high-risk systems must:
– conduct verifications to ensure that these systems are safe, reliable, and legal
– ensure transparency in the operation of the systems,
– comply with data management and cybersecurity requirements,
– provide human supervision.
What is the scope of the ban on social scoring systems?
The Act prohibits the use of social scoring systems that assess the social credibility of individuals. I.e. these systems cannot be used to make decisions about things like access to credit, insurance, or employment.
How does the AI Act support innovation?
The AI Act aims to support development and innovation in the field of artificial intelligence. The European Commission is tasked with developing guidelines and standards for AI systems. It also promotes research and development in this field.
Where can I find more information about the AI Act?
More information about the AI Act can be found on the websites of the European Commission and the European Parliament.