The European Parliament has approved a new bill aimed at making the use of artificial intelligence safer and more transparent.
Artificial intelligence (AI) is a branch of computer science that deals with the creation of intelligent agents, i.e. systems that can reason, learn, and act autonomously. AI research has made significant progress in developing effective methods for solving a wide range of problems.
MEPs have approved new rules and guidelines to ensure ethical and human-centered development of artificial intelligence (AI) in Europe. The Committee on the Internal Market and the Committee on Civil Liberties voted in favor of the draft negotiating mandate with 84 votes in favor, 7 against, and 12 abstentions. The purpose of the amendments introduced by MEPs is to establish rules that guarantee the supervision, security, transparency, traceability, non-discrimination and environmental friendliness of artificial intelligence systems. The definition of AI should be technology-neutral and apply to both current and future AI systems.
The main goal of the AI Law is
One of the leading members of the European Parliament, Romanian Dragos Tudorache, emphasized the importance of this law, saying that it is likely to be the most important piece of legislation in the current parliament. He emphasized that this is the first law of its kind in the world that allows the European Union to take the lead in creating human-centered, reliable, and safe AI practices.
The Regulation takes a risk-based approach and imposes obligations on providers and users depending on the level of risk associated with the AI system. AI systems that pose an unacceptable risk to human safety will be strictly prohibited. This includes systems that use subliminal or manipulative methods, exploit vulnerabilities, or engage in social evaluation based on personal characteristics and behavior.
MEPs have made significant changes to the proposed list, which now includes a ban on the intrusive and discriminatory use of AI systems. Examples of such prohibitions include real-time remote biometric identification systems in public places, except for law enforcement purposes with a court authorization for serious crimes. Biometric categorization based on sensitive characteristics, predictive policing systems, emotion recognition systems in various situations, and indiscriminate biometric data collection for facial recognition databases are also prohibited.
The classification of high-risk areas was expanded by MEPs to cover harm to health, security, fundamental rights, the environment, as well as AI systems that influence voters in political campaigns and recommendation systems used by social media platforms with more than 45 million users.
As for general-purpose AI, providers of basic models such as GPT will be required to ensure that fundamental rights, health, safety, environment, democracy, and the rule of law are well protected. They will have to assess and mitigate risks, comply with design, information, and environmental protection requirements, and will be forced to register in an EU database.
Generative repository models will be subject to additional transparency measures, such as disclosing that content was created using AI, preventing the creation of illegal content, and publishing summaries of copyrighted data used for learning.
To support innovation in the field of AI, exemptions have been added for research activities and artificial intelligence components provided under open source licenses. The new law also promotes the creation of regulatory sandboxes – controlled environments where government agencies can test AI before it is deployed.
The role of the EU Artificial Intelligence Authority
MEPs are seeking to expand the rights of citizens by allowing them to file complaints against AI systems and receive explanations of decisions made by high-risk AI systems that significantly affect their rights. The role of the EU’s AI Authority will be reformed to monitor the implementation of rules in the field of artificial intelligence.
The draft negotiation mandate requires the approval of the entire parliament before negotiations with the EU Council can begin. The vote is expected to take place during the session on June 12-15, 2023.
Categories of artificial intelligence systems
The law defines three categories of artificial intelligence (AI) systems:
- high-risk;
- are unacceptable;
- low-risk.
High-risk AI systems are those that pose a significant risk to human security, fundamental rights, or democracy.
Unacceptable AI systems are those that are considered inherently harmful, such as those used for social scoring or mass surveillance.
Low-risk AI systems are those that do not pose a significant risk and therefore are not subject to the same level of regulation.
The law requires high-risk AI systems to undergo a rigorous assessment process before they can be used. This assessment process takes into account the potential risks and benefits of the system, as well as the measures that have been taken to mitigate the risks. The law also requires that high-risk AI systems be transparent so that users can understand how they work and make informed decisions about use.
The law is a significant step forward in the regulation of AI. It is the first law in the world to set comprehensive boundaries for the development and use of AI. The law is expected to have a significant impact on the development of AI in the EU and is likely to be followed by other countries around the world.
K lucid provisions of the law on artificial intelligence
High-risk AI systems must undergo a rigorous evaluation process before they can be used. This evaluation process takes into account the potential risks and benefits of the system, as well as the measures that have been taken to mitigate the risks.
High-risk AI systems need to be transparent so that users can understand how they work and make informed decisions about whether to use them. This transparency is achieved in various ways, for example, by providing information about the system’s algorithms and datasets.
Companies that develop or use high-risk AI systems must comply with a number of obligations. For example, they must take measures to mitigate risks and ensure responsible use of the systems. Companies that fail to comply with these obligations can be fined up to €20 million or 4% of their global annual turnover.
The AI law is expected to have a significant impact on the development and use of AI in the EU. The AI law is likely to lead to the development of more transparent and responsible AI systems, but may also make it more difficult for companies to develop and use high-risk AI systems that pose a significant risk to human security, fundamental rights, or democracy.
However, the AI Act has faced opposition from and from large US tech companies that dominate the AI field. This could lead to increased transatlantic conflicts over AI technology. As generative AI models such as ChatGPT gain popularity, the global public debate about AI is heating up. While US President Biden’s administration has been reluctant to propose new laws, instead focusing on voluntary cooperation with industry and stricter enforcement of existing laws against fraud and privacy violations, EU legislation has taken the opposite approach.
Attitudes to AI regulation vary across the world, and governments are trying to understand the implications of this technology. The UK has so far taken a relatively restrained approach, while the Italian data protection agency temporarily banned access to ChatGPT in April 2023, but later restored this access .
Expert opinions on the new Law
Experts’ opinions on the new law are divided. Some experts consider the law a necessary and important step towards ensuring the responsible use of AI. They argue that the law’s norms will serve as a safeguard against AI misuse, such as discriminatory practices and social scoring, and will ultimately protect people and society.
On the other hand, , some experts believe that the AI law is overly restrictive and will hinder innovation in the AI sector. They argue that regulation will create additional obstacles for businesses to develop and implement AI products and services, potentially slowing down technological progress.
Timnit Gebru, former co-leader of Google’s AI team, said: “The AI law is a good start, but it is not enough. We need a complete ban on the use of AI for social scoring and other harmful applications. We also need to ensure that AI systems are accountable to the public. The AI Act is a step in the right direction, but it is not enough to protect us from the dangers of AI.”
Clara Montiel, a researcher at the Ada Lovelace Institute, a think tank focused on AI policy, said: “The Law on AI is an important step forward. The AI Law is a landmark piece of legislation that will set a global standard for AI regulation. It is ambitious in scope and covers a wide range of AI systems, from facial recognition to social scoring. The law also contains strong guarantees for fundamental rights, such as privacy and non-discrimination. However, there are some concerns that the AI law may be too burdensome for businesses and may stifle innovation. It is important to find a balance between protecting fundamental rights and ensuring that AI can be used for the benefit of society.”
The true impact of the AI law remains to be seen as the implementation of the AI law and the consequences unfold over time. Nevertheless, it is clear that the AI law is an important milestone in AI regulation. It will undoubtedly impact the use of AI in Europe and can serve as a model for other countries around the world as countries consider their own regulatory boundaries in the field of AI.
artificial intelligence / copyright / European Parliament / intellectual property / new copyright law
