Portugal - National AI Agenda

Overview

The Agenda Nacional de Inteligência Artificial is a draft strategic roadmap promoted by the Portuguese Government under the umbrella of the Estratégia Digital Nacional — Agenda Nacional de IA. Launched into a process of public auscultation in January 2025 and coordinated by the Agência para a Modernização Administrativa (AMA), the Agenda sets an ambition for 2030 to build an AI ecosystem that is ethical, scientifically excellent and economically transformative. It focuses on three pillars — Talent, Innovation & Research and Infrastructure — and lists operational measures including the national LLM project (Amália), participation in a European AI Factory and acquisition of dedicated computing resources to accelerate public and private sector AI adoption. The document is intentionally strategic and consultative: it collects inputs from citizens, business, academia and public bodies to design a plan of actions for 2025–2026 while aligning with EU frameworks.

Definitions

Key terms in the Agenda follow widely used international definitions: "Artificial Intelligence" covers software and systems that perform tasks normally requiring human intelligence, including machine learning, deep learning, and large language models (LLMs); "AI Factory" refers to high‑performance, shared infrastructure and operational services for training and deploying large models (a concept developed at European level via EuroHPC); "Model documentation" refers to structured records about datasets, training processes, evaluation metrics and risk mitigation; "Public sector AI" denotes AI systems developed or procured by governmental entities to support public services. The Agenda emphasizes "ethics", "fundamental rights" and alignment with the EU AI Act and GDPR as baseline governance anchors.

Governance and Institutional Framework

The draft Agenda assigns central coordination to AMA and foresees operational involvement from the Fundação para a Ciência e a Tecnologia (FCT), sectoral ministries (e.g., Health, Finance), research institutions and supervisory authorities. A specialised Committee of Accompaniment — populated by academics, industry representatives and civil society — was established to advise and validate actions; press coverage and AMA communications named distinguished experts as members. The Agenda proposes formal mechanisms for stakeholder engagement (public auscultation sessions, an online contributions phase via Participa.gov.pt) and envisages joint governance bodies to manage public investments, procurement rules and shared infrastructure. It also anticipates coordination with Portugal’s national data protection authority, the CNPD, for compliance with GDPR-related aspects of public sector datasets and model training.

Key Focus Areas

The Agenda organises actions around three interdependent focus areas. First, Talent: national programmes to reskill public sector workers, expand AI curricula in higher education and attract international researchers. Second, Innovation & Research: funding for academic‑industry consortia, support for SMEs to adopt AI, open research initiatives, and the development of a Portuguese LLM (Amália) with open‑source commitments for public utility. Third, Infrastructure: participation in European AI Factories, procurement of dedicated compute, establishment of experimental platforms and data‑sharing frameworks for responsible model training. Cross‑cutting priorities include: embedding ethical by‑design processes, mandatory model documentation for publicly funded models, sectoral pilots (healthcare, justice, public administration) with rigorous evaluation plans, cybersecurity and resilience measures for model integrity, and international collaboration (EuroHPC, EU research programmes). The Agenda also emphasises alignment with the EU AI Act, GDPR, and other relevant frameworks to ensure legal compliance and cross‑border interoperability.

Implementation Framework

Implementation is proposed through a phased Plan of Action (2025–2026) listing concrete measures, responsible entities, indicative budgets and performance indicators. The FCT is identified to coordinate research and infrastructure projects (including the national LLM and AI Factory participation), while AMA leads operationalisation and public sector adoption. The Agenda recommends the creation of funding calls (PRR/COMPETE/FCT), procurement frameworks that integrate ethics and safety criteria, and capacity building for procurement officers to incorporate AI risk assessments. It supports experimental sandboxes for safe public sector testing and requires agents using public funds to produce model documentation and impact assessments. The Agenda recommends integration of model safety testing and conformity assessment steps into procurement and R&D contracts.

Monitoring and Evaluation

Monitoring is to be performed by a central follow‑up committee reporting to the Government and by designated institutional leads for each action. The Agenda specifies outcome indicators (talent pipeline metrics, number of sectoral pilots, compute hours procured, model documentation completion rates) and process indicators (public contributions received, stakeholder engagement events). It also recommends periodic public progress reports and open monitoring dashboards. Evaluation criteria integrate technical performance, compliance with data protection and fundamental‑rights safeguards, economic impact and uptake by SMEs and public administrations.

Penalties, Liability, and Appeals

As a strategy rather than a legally binding regulation, the draft Agenda does not itself set new administrative penalties. It emphasises that enforceable obligations will follow from applicable laws (e.g., GDPR, sectoral regulation) and from the EU AI Act once in force. The Agenda instructs public bodies to ensure contractual liability clauses, audit trails and redress mechanisms for AI systems procured or developed with public funds. It also recommends that model developers and procuring entities maintain documentation to facilitate investigations and administrative reviews by competent authorities, and that grievance/appeals channels be established for affected citizens under existing legal frameworks.

Relationship to Other Instruments

The Agenda is explicitly nested within the Estratégia Digital Nacional and coordinates with national funding instruments (PRR, COMPETE) and research programmes managed by FCT. It declares intent to align with the EU AI Act and the General Data Protection Regulation (GDPR) implemented by the CNPD. The Agenda also references European initiatives such as EuroHPC / AI Factory networks as part of its infrastructure strategy to ensure cross‑border technical interoperability and shared compute resources.

International Alignment

The Agenda positions Portugal within European ecosystems: participation in a EuroHPC AI Factory consortium, collaborative research across EU programmes and alignment with the EU AI Act and Programme for the Digital Decade. It stresses international cooperation for model safety research, standards development and ethical frameworks (e.g., IEEE, ISO initiatives) and highlights co‑funding opportunities with EU instruments. By fostering open research outputs (where appropriate) and interoperability standards, the Agenda aims to ensure Portuguese models and services can integrate into EU cross‑border services and benefit from shared governance approaches.

Implementation Timeline

Milestone	Date
Start of public auscultation and contributions	2025-01-01
Public auscultation sessions (Lisbon, Évora, Porto)	2025-01-15 — 2025-01-17
Online contributions phase opens on Participa	2025-01-30
Deadline for contributions (participa phase)	2025-02-28
Public consultation planned (Consultalex)	March 2025 (planned)
Initial planned Government presentation of Agenda	2025-03-31 (planned; later reported postponed)
Operational Plan (2025–2026) & first funding calls	2025 Q2–Q3 (implementation window)

Compliance Checklist

Requirement	Responsible	Evidence / Action
Model documentation for publicly funded models	Project owners / AMA / FCT	Release of model cards and datasheets; publication on project portals
GDPR compliance for training data	Data controllers / CNPD guidance	Data protection impact assessments; records of processing
Ethics and rights impact assessment	Project leads	Published assessment and mitigation plans
Security and resilience testing	Infrastructure providers / FCT	Penetration & red‑team reports; continuity plans
Procurement with AI risk criteria	Public purchasers	Contracts include audit & liability clauses

Sources and References

Source	Type
Estratégia Digital Nacional — Agenda Nacional de Inteligência Artificial (digital.gov.pt)	Primary Source
AMA — AMA promove sessões públicas para a construção da Agenda Nacional de IA	Primary Source
Agência para a Modernização Administrativa (AMA)	Primary Source