AI sovereignty is the single theme EU heads of state or government will tackle on Friday 13 November, when António Costa gathers the 27 for an informal European Council hosted by the Irish Presidency. The choice is deliberate, because AI sovereignty now shapes debates on budgets, industry and security alike. Costa wants leaders to decide how Europe competes in artificial intelligence before money and rules for the next decade are locked in.
His invitation letter, sent on 3 November with a steering note, frames the meeting as a strategic conversation rather than a decision-making summit. The note concedes that Europe has strong foundations in research, talent and industrial data. It also admits that the Union depends on foreign suppliers at several points of the AI value chain, from chips to cloud capacity and frontier models.
The format reflects that diagnosis. A morning session opens with Arthur Mensch, chief executive of Mistral AI, who will speak on Europe’s position in the global race, on cutting critical dependencies and on getting industry to adopt AI at scale. Leaders then face two guiding questions: how to make Europe a global and sovereign AI player, and how to mobilise public and private investment at the scale a competitive European ecosystem requires. Both questions test what AI sovereignty means in practice.
The second discussion turns to society and security. Geoffrey Hinton, the 2024 Nobel laureate in physics often called a godfather of AI, will address labour markets, democracy, the protection of minors, cybersecurity, market concentration and international governance of advanced systems. Leaders are asked whether the current regulatory framework needs adapting and how the Union should shape global AI rules.
The economic stakes explain why AI sovereignty has moved from technical councils to the leaders’ table. Costa explicitly links the debate to the 2028-2034 multiannual financial framework and to the proposed European Competitiveness Fund, the instrument meant to pool industrial and research spending. Any ambition voiced in Dublin will only matter if it survives the budget fight now under way between capitals.
That tension runs through the whole AI sovereignty exercise. Governments that want faster deployment of AI gigafactories and compute capacity also want tight control over the size of the next EU budget. Net payers will resist open-ended commitments, while smaller economies fear that a sovereignty agenda built around large champions could leave their firms behind.
Regulation adds a second fault line to AI sovereignty. Several capitals argue that the AI Act and related digital rules must stay stable to protect citizens and investors alike. Others push for lighter obligations so European developers can scale before American and Chinese rivals widen their lead. Hinton’s presence ensures that safety, trust and fundamental rights will not disappear from the room.
No formal conclusions are expected, and Costa aims to wrap up by 4 pm after a working lunch among the 27 on the way forward. He has already signalled that AI will return to the agenda in 2027, including at the leaders’ retreat at Alden Biesen in Belgium. Friday is therefore a framing exercise, intended to give the Commission and ministers a clearer political mandate.
For businesses and investors, the signal to watch is whether leaders accept a shared definition of AI sovereignty that balances openness with strategic control. A narrow definition would tilt the Union towards protection and preference rules. A broad one would put investment, skills and access to compute first, and the choice made on Friday will shape both the budget and the rulebook that follow.
