We’re proud to share that OpenMined’s Head of EU Policy, Peter Ide-Kostic, has been selected by the European AI Office to serve on the Expert Forum on Frontier AI.
What is the Expert Forum on Frontier AI?
The European Frontier AI Initiative was announced in the European Commission’s Apply AI Strategy in October 2025 and launched at the Summit on European Digital Sovereignty in Berlin the following month. Its objective is to enhance the competitiveness, sovereignty, and security of the European Union in trustworthy frontier AI.
Following an open call for expressions of interest, the AI Office selected a group of senior experts spanning European AI developers, investors, industry, technical research, academia, and think tanks working alongside representatives from Member States and the Commission to map existing frontier AI efforts, identify bottlenecks and challenges, and shape strategic recommendations for how Europe can access, choose, control, and benefit from frontier AI.
Why this matters for OpenMined’s mission
Europe’s frontier AI questions ranging from sovereignty to security and the ability to genuinely choose among AI capabilities rather than depend on a handful of centralized providers are the questions OpenMined thinks about everyday.
Peter’s selection reflects the growing policy relevance of the ideas our community has spent years developing:
- Attribution-Based Control (ABC) offers a path to unlocking the world’s non-public data for AI, not by weakening control, but by strengthening it. When data owners can govern exactly how their information contributes to a model and receive attribution for it, vastly more of the world’s knowledge becomes available for AI development. For a continent seeking sovereign AI capability without sacrificing its values on privacy and data protection, ABC provides novel solutions.
- Network-sourced AI (NSAI) points toward a future where frontier capability doesn’t have to be concentrated in a few labs. AI built across a network of data owners, each retaining control, is a fundamentally different strategic position than AI built by scraping and centralizing.
- Federated learning (FL) and privacy-enhancing technologies (PETs), embodied in tools like PySyft and SyftBox, make these ideas practical today, allowing institutions to collaborate on AI without surrendering their data.
These are precisely the kinds of approaches Europe will need if it wants frontier AI that is competitive and trustworthy.
What’s next
The first meeting of the expert forum will take place on 23 April 2026. Discussions within the Forum are held under the Chatham House Rule, and the AI Office committed to continue publishing key insights and updates on the Frontier AI Initiative as the work progresses.
We’re proud of Peter for representing OpenMined and the broader movement for privacy-preserving, decentralized AI in this important venue and look forward to hearing more about the forum’s work.