The Davos Declaration on AI Research and Innovation

We,

scientists, innovators, entrepreneurs, policymakers, experts in education, energy, environmental, and labor market issues,

some of us, participants in the Davos AI Horizon25 conference, where the foundation of the declaration was laid, and, above all,

Citizens of Europe and beyond, committed to the responsible advancement of Artificial Intelligence,

• acknowledging that AI constitutes not only a technological disruption but also entails profound economic, societal, and human impacts,
• knowing about the huge opportunities but also the enormous risks associated with AI,
• recognizing that the development, regulation, and deployment of AI are inherently cultural and must be grounded in fundamental values,
• understanding that AI plays a strategic role in global geopolitical competition, both economically and militarily,
• noting that Europe must significantly accelerate its capabilities in AI applications and the development of large-scale datasets,
• considering that research freedom and the mobility of researchers are increasingly constrained in other regions of the world, with a growing influx of talent toward Europe,
• affirming that research is inherently political and must be democratically legitimized,

we hereby declare:

1. Innovation and Research Aligned with European Values

AI research and innovation in Europe shall be deeply rooted in European fundamental rights, ethical principles, and democratic values. The focus should be on creating human-centered AI that boosts human abilities rather than replace them. The objective is to advance inclusive, transparent, secure, and trustworthy AI serving individuals and society alike.

2. Freedom of Research

AI research must stay free and independent. Authorities shall refrain from direct intervention in the scientific process, focusing instead on strategic priorities and enabling frameworks. The outcomes of publicly funded AI research shall be made openly accessible.

3. Openness and International Collaboration

European AI research should be open and actively promote international cooperation. The principle of the free movement of researchers and scientific knowledge must be protected. Europe shall position itself as an attractive, stable, and secure environment for AI researchers and developers, providing legal certainty and a long-term strategic perspective, offering solutions to like-minded countries.

4. Democratic and Bottom-Up Driven Research Policy

AI research policy in Europe should mainly follow a bottom-up approach, fostering creativity, diversity, and responsiveness. By supporting the exploration of new, interdisciplinary, and unexpected research avenues, Europe can ensure that its AI research ecosystem remains dynamic, innovative, and globally competitive.

5. Reproducible and Collaborative Research

Publicly funded projects must be required to publish open code, data, and methods that can be easily reproduced. Europe should invest in common platforms for transparent peer review, collaborative development, and the creation of AI commons.

6. A Unified and Interconnected Research Landscape

To avoid siloed research and improve impact, alternatives to inflexible multi-year programs should be developed. Funding should ensure new projects build on the results of previous ones. Research should be merged with business units in companies with higher technology readiness levels. Calls for funding are expected to prioritize continuity, scalability, and cross-border collaboration instead of duplication.

7. Explainable and Trustworthy AI

To build genuine trust in AI, Europe must lead the co-development of explainable high-risk applications, enabling citizens to meaningfully understand, audit, and contest AI decisions. Innovation funding and research must prioritize sector-specific explainability standards reflecting domain-related risks, regulatory frameworks, stakeholder perspectives, and ethical implications.

8. Self-Sovereign Identity and Data Portability

Europe must establish a robust legal framework for self-sovereign identity that gives citizens control over their data while ensuring portability and cross-border interoperability. Privacy-preserving technologies such as zero-knowledge proofs should be promoted as core components of this infrastructure. Self-Sovereign Identity provisions should be expanded within the General Data Protection Regulation (GDPR), aligning privacy with innovation and enhancing compliance through modern, secure frameworks.

9. Data Access

A balanced approach is needed that protects privacy while giving European companies and researchers access to the rich, diverse, and high-quality data needed to build AI models. Public data partnerships, open training resources, and free AI tool access should be supported.

10. A Harmonized Pro-Innovation Regulatory Framework

The current patchwork of regulations hampering cross-border collaboration and startup scalability amust be replaced by a harmonized European framework, covering certification, reporting, and oversight. Expanding regulatory sandboxes will let innovators test AI in real-life settings, with solid but supportive supervision.

11. A Competitive and Entrepreneurial AI Ecosystem

Europe must move beyond a compliance-led mindset to promoting risk-taking, entrepreneurship, and breakthrough innovation. Support for startups needs to go beyond funding; it should include pilot programs, mentorship, and access to skilled experts to encourage practical experimentation. A Europe-wide fund could back big, ambitious projects – moonshots – that lower the risk for private investors. A pan-European fund should back moonshots and reduce risk for private investors.

12. Embodied and Edge AI: Reviving Europe’s Hardware Legacy

Reviving Europe’s hardware legacy requires targeted investment in edge devices and consumer AI technologies. This includes national and EU-level support for chip design, firmware, and secure operating systems. The future of AI isn’t just in the cloud, it’s in our phones, robots, and wearables.

13. Pan-European Infrastructure for AI Development

Europe needs its own AI infrastructure. A pan-European training facility – akin to CERN – should be created to build multilingual large language models trained on public data. These models should serve public-interest applications in areas like education, healthcare, and law. Instead of competing to develop general-purpose models, Europe should lead in trustworthy, specialized AI aligned with its values.

14. Strategic Use of Europe’s Regulatory Power

Europe’s leadership in AI regulation is a strategic advantage. This should be used to demand data reciprocity. If European data trains other models, Europe must gain proportional access. Making certification and compliance a prerequisite for government contracts could make EU standards a powerful bargaining chip in global negotiations.

15. Decentralized AI for Democratic Autonomy

To promote democratic control over technology, Europe must invest in decentralized AI architectures that align with democratic values and digital sovereignty. Partnerships with like-minded countries should support federated learning, open-source models, and local agents, all designed to favor transparency and control.

16. Labor Empowerment

AI must support labor transitions through inclusive reskilling and upskilling programs, AI-powered and personalized training. Safety nets like unemployment benefits or pilot programs for basic income should be part of the initiative.

17. Education

Ethics and digital literacy should be core components of European education systems from an early age. Public campaigns should raise awareness of user rights, explainability, and safe digital practices.

18. Certification

Developing a harmonized AI Governance Certification Exam is essential to training the next generation of professionals in AI safety, ethics, and compliance. This industry-wide benchmark should be modular, future-proof, aligned with international standards, and tied to public sector hiring and procurement eligibility.

19. Energy-Aware AI

Europe must ensure AI innovation does not come at the cost of environmental degradation. This requires sustained funding for energy-efficient AI models, carbon-aware computing, and full life-cycle assessments of AI deployments. A comprehensive Green AI Initiative should support the development of low-power hardware and encourage public-private collaboration on sustainable AI infrastructure. Establish clear standards and periodic reporting requirements. At the same time, AI’s energy supply must rely on stable, independent sources to ensure resilience and sovereignty.

20. A Cultural Shift

A new generation of entrepreneurs, engineers, artists, and thinkers will be empowered to take intelligent risks. Neurodiverse teams, generalists, and young innovators must be supported. Europe must embrace failure, reduce bureaucracy, and prioritize fast action over perfection. Europe has to shift its mindset – from caution to courage, while upholding European values and remaining alert to black swan events.

We call upon European institutions, member states, research organizations, innovation promoters, industry stakeholders and civil society to jointly uphold these principles, include them in the research and innovation agendas, and contribute to the open, human-centric, inclusive, and sovereign AI in Europe.

© Copyright “Declaration of Davos” – Supported by MoreThanDigital & #bethechange | Privacy Policy