IFIC: Portugal’s Opportunity to Build Meaningful AI Sovereignty
Portugal has a narrow window in which public investment can do more than accelerate AI adoption. Used well, it can help create national capability, keep sensitive business data under European control and develop a generation of engineers able to build — not merely consume — artificial intelligence.
The Instrument for Innovation and Competitiveness (IFIC), backed by Portugal’s Recovery and Resilience Plan, makes €100 million available through the AI for SMEs programme. The amount matters. The design choices behind it matter more.
A strategic choice, not a software-shopping exercise
Across the world, governments and businesses now treat AI infrastructure as a strategic asset. The United States, China, France, Switzerland and the United Arab Emirates are all investing in models, compute and expertise that they can shape and govern. Portugal should not settle for becoming a passive customer of systems designed elsewhere.
If Portuguese SMEs build their most sensitive processes entirely on public cloud models controlled outside Europe, they inherit structural questions about data location, access, confidentiality and legal exposure. Which data leaves the organisation? Where is it stored? Can prompts or business records be reused? What happens when a supplier changes its terms, architecture or security posture?
The point is not that cloud AI is inherently unacceptable. It is that architecture must follow risk. Some use cases are well served by managed services. Others justify European or on-premises models, stronger isolation, open weights or systems that can be independently audited. Funding criteria should reward organisations that make that distinction deliberately.
Why domestic capability matters
Projects such as Amália — a Portuguese-language model developed with local linguistic and cultural context — illustrate what could become possible. The strategic value lies not only in a single model, but in the ecosystem around it: engineers who can evaluate models, organisations able to fine-tune them safely, and infrastructure that keeps sensitive data under accountable control.
A strong IFIC project should therefore leave the organisation with more than a new subscription. It should create internal capability, documented governance, secure deployment patterns and people who understand how the system works.
A practical example
Consider a Portuguese law firm investing in a domain-specific legal AI system. A credible programme would not begin with a chatbot interface. It would begin with information classification, professional-secrecy requirements, lawful access, model evaluation and an architecture that prevents confidential client material from becoming training data.
The investment could include a European or locally controlled model, carefully curated Portuguese legislation and case law, role-based access, human review and a small internal AI engineering team. The return would come through faster research, more consistent knowledge management and greater capacity — while the firm retains control of the data on which its reputation depends.
What funding assessments should ask
Every proposal involving public model APIs should be able to answer a small set of serious questions:
- What information will the system process, and how is that information classified?
- How has prompt injection and data-exfiltration risk been assessed?
- What is the incident-response plan for an AI-related data leak?
- Why is the proposed deployment model proportionate to the risk?
- How will the organisation demonstrate compliance with GDPR, NIS 2 and the EU AI Act?
- What skills and reusable capability will remain when the funded project ends?
Two possible outcomes
IFIC can subsidise dependence, leaving hundreds of companies tied to external platforms with little new knowledge of their own. Or it can create a network of Portuguese organisations able to evaluate, deploy and govern AI responsibly.
Portugal has the talent to choose the second path. The opportunity is to use public funding not simply to purchase AI, but to build a more capable, sovereign and resilient digital economy.
Henrique Necho
CEO, NECHO TECHLAW · Member of the Portuguese Technical Committee for Artificial Intelligence (CT 223)