The European Union’s Path to Digital Sovereignty with the “AI Continent Action Plan”

The European Union (EU) aims to be not only a technological player in artificial intelligence (AI) but also an ethical, human-centred leader. In line with this vision, the “AI Continent Action Plan”, published on 9 April 2025, attracted attention as the latest link in the EU’s long-standing strategy of “trustworthy and human-centric AI”. This plan, however, is not a starting point; it can be seen as the continuation of a vision built step by step since 2018.

The European Model: Balancing Excellence and Trust

The EU’s approach to AI is meant to rest on the concepts of excellence and trust. This vision was first set out clearly in the 2020 “White Paper on AI”:

• Excellence: Under this goal, increasing research capacity and supporting AI laboratories and supercomputing infrastructure stand out as sub-headings.
• Trust: Under this goal, the aim is to develop systems that respect fundamental rights, are consistent with ethical values, and are transparent and auditable, and to support society’s sense of trust in AI applications.

These principles were given a legal basis with the “AI Act”, proposed in 2021 and adopted in 2024, and the EU thus put forward the world’s first comprehensive legal framework dedicated to AI. Prohibited practices have applied since February 2025 and the obligations for general-purpose AI models since August 2025; under the Digital Omnibus regulation published in July 2026, the rules for high-risk systems were postponed to December 2027 (and to August 2028 for those covered by product legislation).

The Digital Decade and the Political-Economic Background

The EU’s “Digital Decade” vision, which sets its targets for 2030, is regarded not only as a technological but also as a political and economic strategy.

In this context, the following programmes stand out:

• Horizon Europe & Digital Europe: These programmes target AI investment of 1 billion euros a year.
• Recovery and Resilience Facility: This programme includes 134 billion euros of funding for the digital transition.
• GenAI4EU: Its main goal is to spread generative AI in the public sector and industry.
• AI Innovation Package (2024): A programme offering incentives to start-ups for ethical and trustworthy AI.

These initiatives also show that the EU aims for Europe to be not only a producer of technology but also a centre of values-based technological development.

“AI Continent Action Plan (2025)”: A New-Generation AI Push

The AI Continent Action Plan marks the EU’s entry into a new era in AI policy. The plan is built around the following five strategic areas:

• AI Factories and AI Gigafactories: Building large-scale AI development infrastructure is seen as one of the plan’s main strategies.
• Data Union Strategy: Through a strategy to make cross-border data sharing easier, organisational coordination is sought.
• Apply AI Strategy: Based on integrating AI into health, public services, science and industry, this strategy aims to increase the use of AI in everyday areas and to strengthen the creation of public benefit.
• AI Skills Academy: In line with the goals of talent development and skilled migration policy, the aim is to strengthen AI research and development and to give the employment of expert staff a systematic infrastructure.
• AI Act Service Desk: As part of support for implementing the legislation on the ground, the aim is to monitor the implementation of the strategic plan within the set legal framework and to develop the legislation in line with emerging needs.

The plan takes on the role of making the EU not just a market but also an ethical centre of technology production. The Apply AI Strategy envisaged in the plan was published in October 2025 and the Data Union Strategy in November 2025.

The Heart of Regulation: The AI Act and the Risk-Based Approach

The AI Act offers a new regulatory model by classifying AI systems into four risk categories:

• Minimal risk: The risk level covering harmless and widespread applications. Example: spam filters used in e-mail services.
• High risk: The risk level covering applications that may affect fundamental rights. Example: AI systems that select candidates in recruitment processes, biometric facial recognition systems.
• Unacceptable risk: The risk level covering systems contrary to fundamental rights. Example: social scoring systems that rate citizens according to their social behaviour.
• Transparency risk: Systems where users must know they are interacting with AI. Example: chatbots that converse like humans or deepfake generators.

In addition, specific rules were introduced for general-purpose AI models (for example large language models): their providers are subject to obligations such as technical documentation, a copyright policy and publishing a summary of training data, and models posing systemic risk face additional evaluation and incident-reporting obligations.

Uncertainties and Risks

Although the EU’s human-centred and ethics-focused approach to AI offers an important model, it is also a fact that some uncertainties remain at the implementation stage. Issues such as the allocation of responsibility, the risk of discrimination, the transparency of algorithms and the protection of individual rights in particular require more clarity and work from a technical and legal perspective. How regulation will operate in areas such as general-purpose AI models and through which mechanisms oversight will be ensured in practice are questions still open to development.

Conclusion: Can the European Model Be a Global Example?

With moves such as the “AI Continent Action Plan” and the “AI Act”, the European Union has succeeded in placing the idea that “AI must work for humanity” within a legal and strategic framework. This vision goes beyond technological progress and also carries a claim to ethical leadership. Sustaining this leadership, however, seems possible not through legislation alone but through approaches that prioritise interdisciplinary coordination and through backing them with practical effectiveness in implementation.