European Artificial Intelligence Act Comes into Force

Landkarte *** Europa | Image credit: ©beugdesign - Stock.adobe.com

On 1 Aug. 2024, the European Artificial Intelligence Act (AI Act) came into force (1). The AI Act is designed to ensure that artificial intelligence (AI) developed and used in the European Union (EU) is trustworthy, and that there are safeguards in place to protect people’s fundamental rights. In January 2024, the European Commission (EC) launched a package of measures to support European small and medium enterprises (SMEs) and start-ups in the development of trustworthy AI, and in May, the EC established an AI Office (2,3). In July 2024, the amended European High Performance Computing Joint Undertaking (EuroHPC JU) Regulation came into force to further develop, deploy, extend, and maintain the EU supercomputing and data infrastructure (4). This paved the way for the establishment of AI Factories, dedicated AI supercomputers, that can be used for training General Purpose AI (GPAI) models (5).

Implication of the AI Act

The AI Act aims to harmonize AI across the EU to encourage the uptake of this technology and support innovation and investment. The AI Act also prohibits certain AI practices and sets out regulations on “high-risk” AI systems and those that pose transparency risks (Figure 1).

Figure 1. Definition of artificial intelligence (AI)-based product safety and risk-based approach. (Figure courtesy of the author)

The AI Act also introduces rules to reduce the systemic risk of systemic GPAI models that are capable of performing a wide range of distinct tasks such as the generation of human-like text (6). The new legislation imposes stricter regulation and fines on tech companies operating generative AI models (i.e., DeepMind and OpenAI will fall under the general-use AI category and would be classed as minimal risk). However, developers of high-risk models would be asked to evaluate models, assess and mitigate systemic risks, conduct adversarial testing, report serious incidents to the EC, ensure cybersecurity, and report on their energy efficiency (6).

The European Federation of Pharmaceutical Industries and Associations (EFPIA) welcomes the AI Act and “believes in the potential of applying AI to deliver benefit for patients, life-science companies, and society” (7). However, it stresses that “the regulatory frameworks governing the use of AI in research, development, and manufacture must be fit-for-purpose, risk-based, non-duplicative, globally aligned, and adequately tailored” (7). When AI-enabled tools are used in pharma R&D, they are exempt from the requirements of the EU AI Act; however, developers and deployers of AI-enabled digital health tools (DHTs) and AI-enabled digital medicine would be impacted by the AI Act, but they would not be considered high risk (8).

Enforcement of the AI rules will be phased in

EU Member States have a year to designate national competent authorities to oversee the application of the AI rules. The Commission’s AI Office will oversee the implementation of the AI Act at the EU level, and be responsible for enforcing the rules for GPAI models (1).

The EC will establish the following three advisory bodies to support the implementation of the rules:

The European Artificial Intelligence Board—to ensure the uniform application of the AI Act across EU Member States and to cooperate between the EC and the Member States.

A scientific panel of independent experts—to offer technical advice and input on enforcement. The panel can issue alerts to the AI Office about risks associated with GPAI models.

An advisory forum—a diverse set of stakeholders to offer guidance to the AI Office (1).

The AI Office has the power to enforce these policies. Companies that do not comply with the rules will be fined 1.5–7% of their global annual turnover depending on the violation (e.g., companies can be fined up to 1.5% for supplying incorrect information) (1).

The AI Act’s regulations will take effect in phases. Rules regarding prohibited practices will apply as of 2 Feb. 2025; obligations on GPAI models will apply as of 2 Aug. 2025; and both transparency obligations and obligations on high-risk AI systems will apply as of 2 Aug. 2026 (9).

AI Pact

In the interim, the EC launched the AI Pact to encourage and support organizations before implementing the AI Act (10). The EC has urged the industry to voluntarily commit to the AI Act requirements before the legal deadline. In November 2023, it obtained responses from more than 550 organizations, and the AI Office initiated the development of the AI Pact, based upon two pillars:

Pillar I—establish a gateway to engage the AI Pact network, encourage the exchange of best practices, and provide practical information on the AI Act implementation process

Pillar II—encourage AI system providers and deployers to prepare early and take actions towards compliance with requirements and obligations set out in the legislation.

The AI Pact has enabled the EC to work with stakeholders to support them in understanding the objectives of the AI Act, to adapt and prepare for the implementation of the AI Act, and to share knowledge and build trust in AI technologies (10).

Meanwhile, the Joint Research Centre continues to undertake independent, evidence-based research to support EU policies and determine the societal impact of the AI Act across Europe.

Shaping of EU’s AI policy

The EU AI Act sends a clear message to the rest of the world that Europe is taking AI very seriously and emphasises the importance of trust, transparency, and accountability. The EC is the first governmental body to create a legal infrastructure on AI and has paved the way for others to emulate and replicate.During the coming months, rules will be phased in and European standards will be refined. Therefore, companies need to work closely with the AI Office and other relevant authorities to ensure the safe deployment of AI solutions that will be beneficial to society and drive future innovation in Europe.

References

1. European Commission. European Artificial Intelligence Act Comes into Force. Press Release. Europa.eu. 1 Aug. 2024.
2. European Commission. Commission Launches AI Innovation Package to Support Artificial Intelligence Startups and SMEs. Press Release. Europa.eu. 24 Jan. 2024.
3. European Commission. Commission Establishes AI Office to Strengthen EU Leadership in Safe and Trustworthy Artificial Intelligence. Press Release. Europa.eu. 29 May 2024.
4. European Commission. Setting Up of AI Factories is Now Possible after EuroHPC Regulation Amendment. Europa.eu. 9 July. 2024.
5. European Commission. AI Factories. Europa.eu (Accessed August 2024).
6. Gilbert, S. The EU Passes the AI Act and its Implications for Digital Medicine Are Unclear. npj Digit. Med. 2024, 7, 135.
7. International Bar Association. The Regulation of Foundation Models in the EU AI Act. www.ibanet.org. 12 Apr. 2024.
8. EFPIA. EFPIA Statement on the Use of AI in the Medicinal Product Lifecycle in the Context of the AI Act. 22 Apr. 2024.
9. Covington. EU Artificial Intelligence Act Published. Blog. Global Policy Watch. 19 July 2024.
10. European Commission. AI Pact. Europa.eu. 1 Aug. 2024.

About the author

Cheryl Barton, PhD,is founder and director of PharmaVision, Pharmavision.co.uk.

Article details

Pharmaceutical Technology Europe
Vol. 36, No. 8
September 2024
Pages: 9-10

Citation

When referring to this article, please cite it as Barton, C. European Artificial Intelligence Act Comes into Force. Pharmaceutical Technology Europe 2024 36 (8).