As artificial intelligence becomes an actor capable of analysing, recommending and now taking action, Europe can no longer separate innovation, sovereignty, security and crisis management.

Dependence on foreign models, infrastructure, components or suppliers is no longer merely an economic issue.

It is becoming a strategic risk that could affect public administrations, companies, critical infrastructure and, more broadly, the continent’s decision-making capacity.

These questions will be central to METRICS, the European Congress on Threats, Risks, Innovation, Crisis and Security.

Europe already knows the price of strategic dependencies. It discovered it through Russian gas, shortages of electronic components, the concentration of certain raw materials in China and its dependence on major American digital platforms.

Artificial intelligence could bring all these vulnerabilities together within a single infrastructure.

For several years, we viewed artificial intelligence as a tool capable of producing text, an image, an analysis or a recommendation. That stage is now being surpassed. New systems can consult data, use software, negotiate a purchase, initiate a payment, select a supplier or modify an industrial process. We no longer ask them merely to answer. We are gradually granting them the right to act.

This development profoundly changes the question of sovereignty. If an artificial agent selects an energy supplier, assesses a loan application or manages infrastructure, whoever controls that agent holds part of the decision-making power. Yet this power may depend on a model trained outside Europe, components produced in Asia, cloud services subject to American law and European data processed by foreign companies.

Power will follow permission

Tomorrow, a company will no longer seek only to convince a consumer. It will also have to convince the artificial agent responsible for advising that consumer.

This shift may alter the economic balance of power. An agent may favour the best-documented services, the most technically accessible suppliers or companies already integrated into its environment. European actors could then become invisible, not because their products are inferior, but because foreign systems do not consider them priority options.

The battle for human attention could thus become a battle for machine permission.

The essential question will therefore be: who defines these agents’ selection criteria, who controls the data used, who can verify their decisions and who can revoke their authorisation to act? An artificial agent should have a verifiable identity, a precise mandate, a limited scope of action and an immediate revocation mechanism. For every important decision, it should be possible to know who authorised it, which data was consulted and which person or institution assumes responsibility.

Machine speed must never eliminate human responsibility.

Different risks depending on the power concerned

Not all foreign dependencies present the same danger. They must be examined precisely, without naivety or conflation.

Russia primarily represents a military, cyber and information threat. According to the European Union Agency for Cybersecurity, Russia-linked groups have been among the most active actors targeting public administrations, diplomatic institutions, digital infrastructure and the defence sector in Europe. Artificial intelligence can extend these operations by automating the creation of false content, impersonation, vulnerability identification and the dissemination of narratives intended to divide European societies. European Union Agency for Cybersecurity https://www.enisa.europa.eu/topics/cyber%2Dthreats/threat%2Dlandscape

China raises another category of risks. Its position in raw materials, telecommunications equipment, batteries and several stages of the electronics supply chain gives it considerable economic leverage. An apparently inexpensive artificial service can create lasting dependence if it controls an organisation’s data, interfaces and uses. This dependence becomes particularly sensitive in public administration, transport, telecommunications, energy and defence.

The United States is an essential ally of Europe. Yet a political alliance does not eliminate industrial and legal risks. A large share of cloud services, advanced components and dominant models depends on American companies. American law may also impose certain obligations on those companies, including when data is stored in Europe. European data protection authorities have already highlighted the difficulties this situation can create for European law. European Data Protection Supervisor https://www.edps.europa.eu/sites/default/files/publication/19%2D07%2D10_edpb_edps_cloudact_annex_en.pdf

The danger therefore does not come only from a hostile country. It can also arise from excessive concentration of technological power in a partner whose economic and political interests will not always coincide with Europe’s.

Sovereignty founded on choices

Sovereignty does not mean that Europe must produce everything alone. No country controls the entire chain linking energy, raw materials, components, data centres, models and applications. Being sovereign means being able to choose the conditions of one’s dependence.

A European company or public administration should know the consequences of the sudden disappearance of its main supplier. It should be able to move its data, change models, retain its software and maintain essential activities. It should also know which decisions must never be entrusted to a system subject to foreign jurisdiction.

This ability to substitute alternatives is currently insufficient. The European Court of Auditors recalls that the Union’s share of global component production capacity had fallen to just 9 per cent in 2020. It also highlights the concentration in Asia of several essential stages, particularly advanced production, assembly and testing. European Court of Auditors https://www.eca.europa.eu/ECAPublications/SR%2D2025%2D12/SR%2D2025%2D12_EN.pdf

Sovereignty must therefore be included in economic calculations. The price of a service is not limited to its subscription. It also includes the cost of an interruption, a contractual change, a diplomatic conflict, an export restriction or an inability to retrieve data.

A cheaper supplier can become extremely expensive if no alternative exists.

Building real European capacity

Europe has significant strengths. It has recognised researchers, a powerful industrial sector, supercomputers, manufacturers of strategic equipment and a market of nearly 450 million citizens. The European Commission plans substantial investments in computing capacity, including artificial intelligence factories and several very large infrastructures. These projects can strengthen the ability of businesses, researchers and public administrations to develop models on European soil. European Commission https://commission.europa.eu/topics/competitiveness/ai%2Dcontinent_en

But infrastructure will not be enough. Europe must also encourage public procurement that gives European solutions a place, support open models, require data portability and develop common standards for the identity of artificial agents.

It must also protect certain domains. Military, judicial, energy and democratic decisions cannot depend entirely on opaque systems controlled from abroad. In these sectors, control of data, computing, models and oversight mechanisms is a security requirement.

Deciding before machines decide

Artificial intelligence can deliver scientific discoveries, improve public services, accelerate the energy transition and strengthen European competitiveness. Its ability to produce unexpected decisions is precisely what makes it valuable.

But this same ability can become dangerous when it operates within an electricity grid, a financial system, an election campaign, a cyber operation or a military decision. The question is therefore not whether machines will have greater autonomy. They will.

The real question is whether Europe will retain the authority needed to define their mandate, understand their dependencies and stop their action.

European sovereignty will not mean closing the door to foreign technologies. It will mean never handing the keys to our economies, our data and our democracies to a system we cannot audit, replace or stop.

Several programme round tables directly extend the reflections developed above:

The session “AI Science, Autonomous Agents and Governance: From Fundamental Research to Competitive Advantage” will address the control and accountability of autonomous systems.

The round table “Allies Today, Adversaries Tomorrow: When Our Partners Become Our Vulnerabilities” will analyse strategic dependencies that can become instruments of pressure.

The session “Cyber Research, Complex Systems and Critical Infrastructure: Anticipating Attacks and Designing Tomorrow’s Resilience” will address the protection of essential functions.

Finally, “Industrial Research and Development, Disruptive Technologies and Factories of the Future: Reindustrialising and Building European Sovereignty” will place these issues in the context of technological and industrial control.

METRICS will also address risk forecasting and management through the round tables “Network Science, Predictive AI and Digital Twins: Anticipating Disruptions in Critical Supply Chains”,

as well as “Decision Sciences, Polycrisis Foresight and Leadership: Making Trade-offs under Uncertainty and Preparing Possible Futures” and “Data Science, AI, Open-Source Intelligence and Strategic Foresight: Turning Information into Anticipatory Decisions”.

The aim will therefore not only be to understand threats, but to determine how to detect them earlier, measure their cascading effects, decide more quickly and preserve the continuity of essential activities. Official METRICS programme https://www.metrics-summit.org/fr

Conclusion

European sovereignty in the face of artificial intelligence cannot be limited to political ambition or technological preference. It must become a measurable capability: detecting a threat, understanding dependencies, retaining multiple options, deciding under pressure, maintaining essential functions and rebuilding after a crisis.

This is precisely METRICS’ ambition. Its manifesto states that this gathering is not intended to become just another congress among many. METRICS aims to establish the first milestone towards an independent European framework dedicated to forecasting, crisis management and reconstruction. The objective is to develop a common language, comparable indicators and methods usable by companies, institutions, insurers, researchers, investors and infrastructure operators.

This approach also runs through sessions such as “What Does a Crisis Really Cost? The Hidden Price of Threats, Disruptions and Inaction”, “Risk Science, Predictive Modelling and Insurance Innovation: Financing and Preventing Future Crises” and “Collateral Crises: When the Response to One Crisis Creates the Next”. It will extend into operational simulations devoted to cyberattacks, supply disruptions, reputational crises and decisions made under pressure.

METRICS thus intends to cover the entire crisis cycle.

Before disruption, weak signals must be identified and scenarios anticipated. During a crisis, decisions must be made, coordination ensured and vital activities maintained. After a crisis, essential functions must be restored, recovery financed, lessons drawn from the event and revealed vulnerabilities corrected.

In a context where artificial intelligence, geopolitical tensions, cyberattacks and industrial dependencies can accelerate the spread of crises, Europe needs more than a forum for discussion. It needs a shared memory, a permanent method and a collective capacity for action.

METRICS seeks to help build this framework so that Europe no longer merely reacts to crises, but becomes a true power of anticipation, decision-making and reconstruction. METRICS founding manifesto https://www.metrics-summit.org/fr/manifesto