Advancing Responsible AI Innovation A Playbook 2025

Government leaders CASE STUDY 1 Telefónica’s multi-pronged responsible AI strategy Large organizations often struggle to implement enterprise- wide responsible AI programmes. Rather than starting from scratch, Telefónica, a global telecommunications company, began by building upon its existing risk-oriented privacy governance model and participating in EU-wide efforts to identify AI requirements. After defining a set of AI principles, Telefónica piloted an AI Office, Ethics Expert Group and Responsible AI Champions to steward adoption across teams. The company also tested product and service evaluation methodologies and initiated tailored training and awareness-raising efforts.8 For cohesive and accountable adoption, the company established a unified AI governance framework that integrated compliance requirements alongside ethics to evaluate systems for societal impact, human agency and inclusivity. Key insight Scaling responsible AI requires formalizing principles into a governance model that empowers teams, proactively manages risk, engages cross-functional expertise and supports compliance across diverse regulatory environments. Key roadblocks organizations encounter from the broader ecosystem Sudden regulatory changes and limited guidance, forcing companies to frequently modify their visions, which erodes the consistency of their strategies Increased geopolitical competition and insufficient international AI governance cooperation, making companies choose between geopolitical demands for rapid deployment and responsible AI practices Actions for government leaders –Communicate jurisdictional responsible AI goals: National approaches and expectations regarding governance and regulation must be clearly articulated to encourage organizations to follow suit in communicating responsible AI practices to their employees. Jurisdictions use one or more approaches to communicate goals. Examples include: –National AI strategy: Brazil’s AI Plan (PBIA),9 the US AI Action Plan,10 China’s AI Action Plan11 and Costa Rica’s National AI Strategy12 –Codes of conduct: Canada’s Voluntary Code of Conduct on the Responsible Development and Management of Advanced Generative AI Systems13 and the Hiroshima AI Process Code of Conduct (see Case study 7) –Guidelines: Egypt’s Charter for Responsible AI14 and Australia’s Voluntary AI Safety Standard15 –Regulation: The European Union’s (EU) AI Act, South Korea’s AI Framework Act, and Japan’s AI Promotion Act showcase three divergent approaches to comprehensive AI regulation16Enabling confident adoption by industry requires balancing the frequency of goal revisions by governments with predictability. Governments’ own implementation of robust responsible AI practices can help set adoption expectations for industry. –Incentivize industry implementation of responsible AI: Jurisdictional incentives can help ensure market goals maintain alignment with public interest17 and that an organization’s responsible AI goals also address macro- level challenges like workforce, environmental and information ecosystem impacts. Jurisdictions are exploring varied incentive approaches, such as: –Preferred procurement: such as for AI developers ensuring appropriate guardrails in their models18 –Financial penalties or rewards: including tax incentives, grants or subsidies19 –Standardized frameworks: incorporating expert-informed risk management approaches and reporting templates for responsible AI practices (see Play 6) –Publicity: recognizing companies that meet the jurisdiction’s communicated goals (see Case study 2) Advancing Responsible AI Innovation: A Playbook 10