The AI Act: A Leap Towards Regulating Artificial Intelligence in the EU
The European Union (EU) is making significant strides towards regulating Artificial Intelligence (AI) with its proposed AI Act. This legislation aims to ensure that AI systems used within the EU are safe and respect fundamental rights and EU values.
A Risk-Based Approach
The AI Act follows a ‘risk-based’ approach, meaning the higher the risk an AI system poses to society, the stricter the rules it must adhere to. This approach is designed to mitigate potential harm while fostering innovation and growth in the AI sector.
The Adoption Process
The adoption process of the AI Act is an ongoing journey. The European Commission first proposed the AI Act in April 2021. The European Parliament voted on its position in June 2023, and negotiations to finalize the new legislation began. A provisional agreement was reached on 9th December 2023, marking a significant milestone in the Act’s journey.
In February 2024, representatives from the EU member states formally approved the final text of the AI Act. The Act was then put to a vote by the EU Parliament’s Committee on Civil Liberties, Justice and Home Affairs (LIBE) and Committee on the Internal Market and Consumer Protection (IMCO), who approved the Act with an overwhelmingly favourable vote. This put the Act on track to become law.
The final parliamentary approval of the AI Act is expected in April. Once approved, the Act will enter into force two years after its final approval.
What’s Agreed Upon?
The provisional agreement on the AI Act includes several key elements:
High-Impact General-Purpose AI Models: The Act emphasizes the regulation of high-impact general-purpose AI models that can cause systemic risk in the future. This includes AI systems that have the potential to significantly influence the physical or virtual environment and society at large.
High-Risk AI Systems: The Act also focuses on high-risk AI systems. These are AI applications that have a high potential for harm, such as those used in critical infrastructure, education, employment, essential private and public services, law enforcement, migration, asylum and border control, and administration of justice.
Revised System of Governance: A revised system of governance with some enforcement powers at the EU level has been agreed upon. This is designed to ensure compliance with the Act and to handle any violations effectively.
Transparency and Stress-Testing: The Act introduces dedicated rules for general-purpose AI models to ensure transparency along the value chain. For very powerful models that could pose systemic risks, there will be additional binding obligations related to managing risks and monitoring serious incidents, performing model evaluation and adversarial testing.[1]
Boosting AI Innovation: Beyond governance and enforcement, the AI Act aims to boost AI innovation within the EU. It includes provisions to foster a single market for AI applications, aligning with the EU’s coordinated plan on artificial intelligence to accelerate investment in AI across Europe.[2]
These elements represent a comprehensive approach to regulating AI, balancing the need for safety and ethical considerations with the desire to foster innovation and growth in the AI sector. The final text of the AI Act, expected to be published in the first quarter of 2024, will provide more detailed information on these and other aspects of the legislation.
The Road Ahead
The AI Act is now awaiting sign-off from EU lawmakers before the rules enter into force. This legislation represents a significant step towards a regulated AI landscape in the EU, balancing the need for innovation with the importance of safety and ethical considerations.
Please note that this information is accurate as of the latest available sources and may be subject to change. For the most current information, please refer to official EU resources.
Author: Microsoft. (2024). Copilot. https://copilot.microsoft.com/
As this is the first article and translation (article was written in English language, Slovene version represents a translation), published on our web page which is generated by AI system, I would like to add a legal assessment of authorship and confirm that the essential ethical checks for publication of an article generated by AI system were carried out.
Legal assessment of authorship:
The above article and its translation which is an output of AI system Copilot is considered AI-assisted output (and not AI-generated output), because the text was generated with material human intervention (and not without any human intervention). Only AI-assisted output can classify as “works” under the EU copyright law pursuant to the general copyright principle where only human beings could be authors of works.
Further, the authorship of the above article could be considered only if the article could be qualified as “work” protected under the EU copyright law. For an AI production to qualify as a “work” protected under the EU copyright law the criteria of a “four-step test” established in the report Trends and developments in artificial intelligence[3] prepared for the European Commission shall be met:
Step 1 – Production in literary, scientific or artistic domain – the above article in literary form created as AI-assisted output meets this material prerequisite for production thus we can confirm that this criteria is met;
Step 2 – Human intellectual effort – as the article was generated through AI system (Copilot) which supervised and drove the creative process of the above article and this AI system is a product of human intellectual effort we can confirm that this criteria is met;
Step 3 – Originality/creativity (creative choice) – we can confirm that the AI system was able to express its creative abilities (conception, execution and redaction) in the production of the work by making free and creative choices thus we can confirm that this criteria is met;
Step 4 – Expression – as the output stays within the ambit of the AI system’s general authorial intent, this criteria is also met.
We can conclude that the above article could be classified as “works” under the EU copyright law thus authorship could be established.
As the Copilot has prepared the above article only based on my inserted prompt (“Can you prepare a short article about the AI Act and the current state of the adoption procedure. The article should be short and informative.”), my role in context of authorship is passive because I could not exercise free choices at any stage of the creative process. This means that the developer of the AI system, who has a dominant role at the execution stage, is considered as the author of the above article pursuant to the general copyright principle »copyright vests initially in the person having created the work«. As an author is a natural person who created an author’s work, I rely and trust Microsoft that they have successfully obtained all economic rights on the AI system which was developed by natural person(s). Further the Copilot AI Experiences Terms define that Microsoft does not claim ownership of the Creations. Regardless of this I conclude that it is right to claim that Microsoft is the author of this article, especially as my role in creative process was limited and it is difficult for me to claim that I created the article. Further analysis in respect of this judgement shall be carried out (maybe in a follow up article?), which somehow exceeds purpose of this legal assessment.
Ethical check for publishing AI-generated article on the webpage
Quality control: The article was reviewed and checked for errors by Aljaž Jadek, who confirms that the quality of the article is sufficient for publication and there are no errors in the article (i.e. AI hallucinations).
Avoid Plagiarism: The article was reviewed and checked for plagiarism by Aljaž Jadek. Only sections marked with footnote references are citations from other publications, thus the article does not represent plagiarism.
Maintain Creativity: According to Aljaž’s opinion the content is unique, informative, and engaging and thus suitable for publication.
Ethical Considerations: According to Aljaž’s opinion the content is not misleading or otherwise ethically problematic and thus suitable for publication.
Lebel AI Content: It is transparently disclosed that the article was generated by AI system – Copilot.
SEO Best Practices: The article was not created and published for manipulation of search rankings but for informative purposes and description of the current status of the AI Act adoption procedure.
[1] https://ireland.representation.ec.europa.eu/news-and-events/news/commission-welcomes-political-agreement-artificial-intelligence-act-2023-12-09_en
[2] https://www.eeas.europa.eu/delegations/australia/world%E2%80%99s-first-ai-law-eu-announces-provisional-agreement-ai-act_en?s=163
[3] https://op.europa.eu/en/publication-detail/-/publication/394345a1-2ecf-11eb-b27b-01aa75ed71a1/language-en
