i2CAT Foundation

Get everyone pulling in the same direction on AI by building a shared vision—clearly defining the challenges ahead, agreeing on a practical roadmap, and aligning on common standards. - Establishment of universal human-centric standards: moving beyond technical efficiency to prioritize human dignity, democracy, and the rule of law. - Commitment to digital sovereignty: enabling regions to develop their own AI capabilities and infrastructures (e.g., sovereign clouds, open-source models) to avoid global "algorithmic gaps". - Validation of Key Value Indicators (KVIs): moving beyond traditional KPIs to adopt KVIs that measure social impact, environmental sustainability, and the reduction of the digital divide. - A framework for interoperability: creating mechanisms that allow diverse regional governance models (like the EU AI Act) to work together, ensuring a fair global digital market.

• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Interoperability of governance approaches
• Protection and promotion of human rights
• Transparency, accountability, and human oversight

These choices reflect Catalonia's commitment to pursuing the European approach to AI, one that prioritizes a human-centered framework grounded in ethics, trust, and the protection of fundamental rights: - We prioritize the social pull approach, involving society in tech design rather than just "tech push". Linguistic diversity (e.g., the Aina project in Catalonia) is critical to prevent cultural erosion by global AI models. - Interoperability: i2CAT focuses on Interoperable Data Spaces and federated infrastructures that allow secure data sharing across borders while maintaining local governance - Human rights & transparency: our research emphasizes digital rights and responsibilities. We advocate for public registries of AI systems to ensure the public can trace and audit the algorithms affecting their lives.

Beyond the established themes, we identify several critical emerging issues that require urgent global governance to ensure a resilient and democratic AI global governance in the future: - Digital democracy: AI governance must go beyond "human oversight" to actively foster digital democracy and mitigate risks to democratic quality. This involves addressing how AI, particularly through social networks and generative AI, can threaten election integrity, generate social polarization, and compromise the truthfulness of information through disinformation and manipulation. Furthermore, we must address cognitive dependence, where AI's capacity for automation and cognitive augmentation alters interpersonal relationships and human agency. - Acceleration of technological supremacy and global dependencies: the intense competition for AI leadership-primarily between the USA and China-is creating a new global inequality: the algorithmic gap. With over 80% of digital technologies used in Europe currently imported and 70% of foundation models originating in the United States, there is a systemic vulnerability and worldwide lack of digital sovereignty. - Distributed infrastructure for sovereignty: to counter this supremacy, governance must support decentralized, local infrastructures such as sovereign public clouds, edge computing, and sovereign AI (e.g. https://eurostack.eu/). This allows regions to develop their own competitive technological bases, ensuring control over their digital destiny and lowering external dependencies in critical sectors. - Dual-use security and critical resilience: governance must explicitly address the protection of essential services against cyber-physical attacks. - Sustainability-oriented governance: the massive energy and resource that AI demands, implies that we need the ecological transition to be a core pillar of governance. This mandates the development of sustainable digital infrastructures and the use of AI to solve environmental challenges, such as decarbonization. - New indicators of success (KVIs): current models rely on traditional economic KPIs. We propose adopting Key Value Indicators (KVIs) to quantify the social, environmental, and ethical impact of technology. These metrics-measuring factors like the reduction of the digital divide and the protection of citizen trust-should be used to guide public policy.

Based on the Catalonia AI 2030 Strategy and i2CAT's Global Research and Innovation Roadmap (GRIR), governance gaps and regional advances are identified across the four thematic areas: 1) Social, economic, ethical, cultural, linguistic and technical Implications - Gaps: a critical "algorithmic gap" exists due to unequal access to infrastructure and talent, risking power concentration in a few global actors. Additionally, a massive talent shortage persists, with 34,000+ vacancies vs. 8,000 graduates annually. - Advances: Catalonia is implementing a "social pull" approach to design technology based on societal needs. Key initiatives include Project Aina to protect linguistic diversity, the GovTech project that identifies government needs that can be solved with AI solutions, as well the research on the use of Key Value Indicators (KVIs) to measure social, ethical and environmental impact. 2) Interoperability of governance approaches - Gaps: European digital infrastructure is fragmented, with low interoperability in cloud services and data spaces. Guidelines are often too technical for SMEs to adopt effectively. - Advances: the strategy promotes federated sectoral Data Spaces in health, agrifood, and mobility aligned with European standards. i2CAT is developing "Federated telecom networks as a platform" to unify distributed infrastructure. 3) Protection and promotion of human rights - Gaps: AI poses risks to democratic quality through disinformation and social polarization. There is a lack of binding international legal frameworks for AI-specific human rights. - Advances: the Catalan Charter for Digital Rights serves as a pioneering legal framework. Research also focuses on secure sovereign digital identities using blockchain to protect citizen data. Also initiatives regarding AI literacy for citizens and public workers, that focuses on the ethical and safe use of the technology.

The AI Dialogue should act as a global Policy Lab, bridging ethical theory and human-centric implementation by managing the full policy life cycle: formulation, adoption, and evaluation. Its primary role should be to: - Harmonize governance: develop shared definitions of IA global governance with a human centric perspective, ensuring AI remains aligned with human rights and the rule of law. - Enable cooperation: facilitate international partnerships and the exchange of best practices to avoid fragmented governance. - Establish metrics and standards: create measurable policy recommendations, including the use of Key Value Indicators (KVIs), to monitor global adoption and social impact. - Oversee industry: track global developments to ensure alignment with societal and environmental standards regarding digital rights. - Citizens perspectives: shift to a "social-pull" approach that prioritizes civil society's needs and ensures active participation through co-creation.

The AI Dialogue should connect fragmented regional efforts into a cohesive "Labs of Labs" network, building upon established European and global frameworks: European and other regions regulatory bodies: connect with the European AI Office, the EU AI Act implementation mechanisms, and the AI Safety Institute. - Thematic global dialogues: integrate insights and generate synergies from and with the Vienna Manifesto on Digital Humanism, the Paris AI Action Summit, and the AI for Good Global Summit. - Human rights advocacy: partner with cross-disciplinary networks like Access Now, AlgorithmWatch, XNET, and the European Digital Rights Network to ensure a rights-based foundation. - Research hubs: leverage world-class institutions such as the Barcelona Supercomputing Center (BSC), i2CAT, ELLIS, and the Observatory of Ethics in AI (OEIAC). - Other strategic alliances: it can build on initiatives such as the Disruptive & Emerging Technology Alliance (DETA)—a global open policy laboratory connecting leading tech regions—and the European Network of Living Labs (ENoLL).

We recommend a quadruple helix (4H) (European Network of Living Labs) structure to ensure dialogue and co-creation across all levels of society: - Structure: -----> Government: focus on policy-making and regulatory frameworks. -----> Industry: ICT companies and startups driving digital innovation. -----> Academia: universities and research centres providing cross-disciplinary research and digital ethics and global IA standards. -----> Society: NGOs, user communities, and advocacy groups ensuring public and citizen engagement. - Format: the Dialogue should use Policy Living Labs approach (e.g. https://policy-lab.ec.europa.eu/index_en). To support these initiatives, various collaborative tools—including co-creation workshops and interdisciplinary focus groups—should be employed to integrate diverse insights across technical, sociological, and legal domains, ensuring that AI global governance is built from the bottom up. This approach should be complemented by the implementation of Policy Sandboxes, providing controlled venues to test regulatory impacts and ensure compliance with frameworks such as the EU AI Act before solutions or policies enter the global market.

True global governance must prioritize those currently on the wrong side of the algorithmic knowledge gap. Governance frameworks must adopt an intersectional lens, recognizing that individuals belong to multiple groups simultaneously. This approach ensures that policies address the unique, overlapping systemic barriers and discriminatory patterns, such as the combination of linguistic exclusion, gender bias, and socioeconomic status, that multifaceted identities face in the digital era. These groups should be embedded in the core Policy Living Lab methodology, where participatory design ensures that digital rights and ethical frameworks are built from the bottom up. Specific voices that should be included are: - Linguistic and minority communities: inclusion is vital for non-global languages through initiatives like Aina in Catalonia to prevent cultural erosion and ensure AI reflects regional identity. - Digitally excluded and vulnerable populations: targeted strategies are needed for groups such as the elderly, migrants, and rural populations to overcome the digital divide and ensure equitable access to AI benefits. - Gender and LGTBI perspectives: governance must actively integrate a gender perspective and ensure diverse participation to mitigate algorithmic biases and prevent technology from exacerbating existing inequalities. - Small and Medium-Sized Enterprises (SMEs): small businesses often lack the resources to adapt to complex regulations and find it difficult to source specialized AI talent; they require dedicated diagnostic and support services to remain competitive.

Below some proposals of innovative engagement formats: - Quadruple helix co-creation: use workshops and design-thinking sessions involving government, industry, academia, and civil society to jointly define AI policies reflecting shared values. - Strategic panels: hosting high-level public forums to ensure knowledge sharing. - Expert advisory boards: permanent committees of leading experts to lead policy processes in critical areas like socio-digital inclusion and human-centric governance. - Interdisciplinary focus groups: gather qualitative insights on ethical, sociological, and legal and technological challenges, specifically including vulnerable and underrepresented perspectives. - Matchmaking: facilitate structured networking to accelerate the adoption of responsible AI. - Regional lab networks: a "Labs of Labs" format to coordinate territorial innovation actors and facilitate real-time knowledge transfer across hubs. - Participatory democracy platforms: integrate open-source digital tools like Decidim (https://www.decidim.barcelona/) to facilitate large-scale citizen participation, ensuring the policymaking process is transparent, inclusive, and community-driven.

Catalonia offers several pioneering examples of AI governance that bridge the gap between high-level ethics and practical implementation (https://politiquesdigitals.gencat.cat/web/ca/economia/ia/ok.Eng_Estrategia-IA-Catalunya_VA4_ENG.pdf): - Catalan Charter for Digital Rights and Responsibilities: a landmark legal and ethical framework that ensures technology serves human dignity, democracy, and the state of law. - Public Registry of AI Systems: this platform ensures the transparency and traceability of AI projects used by public administrations, acting as a critical instrument for ethical oversight and accountability. - The PIO Model (Principles, Indicators, and Observables): developed by the Observatory of Ethics in Artificial Intelligence of Catalonia (OEIAC), this self-assessment system provides clear metrics for organizations to align their AI systems with ethical and legal standards. - GovTech Catalunya is the Catalan government's programme that connects the public administration with the Catalan innovation ecosystem, of start-ups (startups and scaleups) and small and medium-sized enterprises (SMEs) that work with disruptive technologies, such as AI, to improve public services and make them more efficient, agile, proactive and adapted to the needs of citizens. - Project Aina: a strategic initiative to develop high-quality open-source Catalan language corpora and computer models, ensuring linguistic sovereignty and cultural diversity in the digital era. - Key Value Indicators (KVI) framework: an innovative methodology used to quantify the social, environmental, and ethical impact of technology-such as digital divide reduction-complementing traditional economic KPIs. - More digitals: a capacity-building program designed to bridge the digital divide by strengthening citizens' core digital competencies. It focuses on essential digital skills, artificial intelligence, cybersecurity, and digital parenting, while promoting the safe, responsible, and informed use of the internet, mobile devices, and emerging technologies. - FRIA Model: promoted by the Catalan Data Protection Authority (APDCAT), this tool facilitates impact assessments on the fundamental rights of individuals affected by AI systems, ensuring compliance with the European AI Act. - CIDAI and DCA-IA: specialized centers and communities that promote the adoption of responsible AI in SMEs through knowledge transfer and innovation challenges.