European AI Act comes into force from August 1, 2024
World’s first comprehensive regulatory framework for AI
Today, August 1, 2024, marks a historic milestone as the European Artificial Intelligence Act (AI Act) officially comes into force, establishing the world’s first comprehensive regulatory framework for artificial intelligence.
The landmark legislation by the European Commission ensures that AI technologies developed and utilised within the European Union are safe, trustworthy, and aligned with fundamental rights. It fosters a balanced approach to innovation and risk management, providing a solid foundation for the future of AI in the EU.
The revolutionary AI Act introduces a nuanced approach to AI regulation, categorising systems based on risk levels.
4 distinct categories:
- Minimal risk: AI systems such as recommendation engines and spam filters fall into this category. Given their low risk to individual rights and safety, these systems face no mandatory obligations under the Act. However, the Act goes further, proactively encouraging companies to adopt voluntary codes of conduct. This initiative is expected to enhance transparency and ethical practices, potentially positively impacting the industry.
- Specific transparency risk: AI systems like chatbots must inform users that they are interacting with a machine. AI-generated content, including deep fakes, must be labelled to indicate its artificial origin, and users must be aware of biometric categorisation or emotion recognition systems in use. Providers must ensure that synthetic content is marked in a machine-readable format to prevent deception. This category sets clear guidelines for AI systems requiring transparency in user interactions.
- High risk: High-risk AI systems, such as those used for recruitment, credit scoring, or autonomous robots, are subject to stringent requirements. These include rigorous data quality standards, risk-mitigation strategies, comprehensive documentation, human oversight, and robust cybersecurity measures. Importantly, regulatory sandboxes will play a crucial role in supporting innovation while ensuring compliance.
- Unacceptable risk: The Act outright bans AI systems that threaten fundamental rights. This includes technologies manipulating human behaviour, such as voice-assisted toys encouraging dangerous actions, ‘social scoring’ systems, and specific predictive policing applications. The Act’s ban on such systems provides a strong sense of security about its protective measures.
The Act also addresses general-purpose AI models—highly versatile systems capable of performing various tasks. These models, often components of different AI applications, will be subject to transparency requirements along the entire value chain. The goal is to mitigate systemic risks associated with these robust AI systems.
Margrethe Vestager, Executive Vice President of Europe Fit for the Digital Age, EU, said: “AI has the potential to transform our lives and economy. The European approach to technology prioritises human rights and sets a global standard for AI development.”
Thierry Breton, Commissioner for Internal Market, EU, added: “Today signifies a major advancement in Europe’s leadership in trustworthy AI. The AI Act delivers a balanced, world-first framework for AI, addressing risks while fostering innovation.”
The AI Act’s implementation will be overseen by the authorities of European countries, which must be designated by August 2, 2025. The European Commission’s AI Office will lead enforcement at the EU level, focusing mainly on general-purpose AI models.
3 advisory bodies to support Act’s rollout:
- European Artificial Intelligence Board: This board will ensure consistent application of the AI Act across EU Member States and facilitate cooperation between the Commission and national authorities.
- Scientific panel of independent experts: This panel will provide technical advice and raise alerts about risks associated with general-purpose AI models.
- Advisory forum: This forum, comprising a diverse group of stakeholders, will offer guidance on various aspects of the Act.
Non-compliance with the AI Act can result in significant penalties. Companies violating the ban on unacceptable AI applications may face fines of up to 7% of their global annual turnover.
Other breaches, such as failing to meet transparency or documentation requirements, could incur fines up to 3% of turnover, with incorrect information penalties reaching 1.5%.
While most of the AI Act’s provisions will be enforced from August 2, 2026, specific prohibitions on high-risk AI systems and rules for general-purpose AI models will take effect sooner—six and 12 months, respectively. To ease the transition, the European Commission has introduced the AI Pact, inviting AI developers to voluntarily adhere to essential obligations ahead of the mandatory deadlines.
In preparation for the Act’s full implementation, the European Commission is developing detailed guidelines and co-regulatory instruments, such as standards and codes of practice. A call for expressions of interest has been opened for drafting the first general-purpose AI Code of Practice, and a multi-stakeholder consultation will allow stakeholders to contribute to the initial Code.
The European Commission has also taken several preparatory steps: on December 9, 2023, it welcomed the political agreement on the AI Act. On 24 January 2024, it launched measures to support European startups and SMEs in developing trustworthy AI.
The European Commission unveiled the AI Office on May 2, 2024, and on July 9, 2024, the EuroHPC JU Regulation was amended to facilitate AI supercomputers for training general-purpose AI models. The Joint Research Centre (JRC) continues to conduct critical research that shapes and implements EU AI policies.
As the AI Act takes effect, Europe has set a precedent for global AI governance, emphasising safety, transparency, and respect for fundamental rights.
Featured image: Most of the AI Act’s provisions will be enforced from August 2, 2026. Credit: Alexandre Lallemand
Last Updated on 5 months by Arnold Pinto
