Advancing Responsible AI Innovation A Playbook 2025

Pre-proposal Bolsters policy adaptiveness Example Open Loop, a Meta-led initiative, prototyped a framework aimed at informing discussions in the EU regarding a potential risk-based approach to AIPost-enactment Address ambiguities and enforcement gaps Example The General-Purpose AI Code of Practice, developed through engagement with over 1,000 stakeholders, is designed to help industry comply with the EU AI Act’s rules on general-purpose AI Example Canada's financial regulator, the Office of the Superintendent of Financial Institutions (OSFI) consulted stakeholders on revising AI model risk guidance, with feedback suggesting only scope adjustments and clarifications were needed, thus avoiding major changes and prototyping challenges Post-enforcement Examines potential material changes via ongoing efficacy monitoring mechanismsExample In the US, cost-benefit analysis of regulations is required as part of the federal rulemaking process, enabling policy-makers to consider the best policy alternative Pre-enactment Identifies unintended consequences and operational needsWhen to prototype policies and regulatory frameworks FIGURE 4 Sources: Meta Open Loop. (2021). AI Impact Assessment: A Policy Prototyping Experiment; US Library of Congress. (2024). Cost-Benefit Analysis in Federal Agency Rulemaking. https://www.congress.gov/crs-product/IF12058#; European Commission. (2025). General- Purpose AI Code of Practice now available. https://ec.europa.eu/commission/presscorner/detail/en/ip_25 1787. –Provide foundational support to enable organizations in responsible AI: Some approaches could include: –Resources and tooling: Singapore provides governance frameworks, tools, training and certifications, such as the Model AI Governance Framework and AI Verify Foundation. –Ecosystem champions: Canada has established three non-profit AI Institutes (Amii, Mila and the Vector Institute) that serve as third-party facilitators of cross-sector interaction and public-private engagement to align priorities and share best practices. –Sandboxes: The UK’s Financial Conduct Authority (FCA) offers services for safe experimentation to firms in various stages of AI, from discovery to use.39 Firms gain regulatory support and validation for confident adoption, while the FCA gains practical insights to shape future oversight and policy. Another example is the Government of India’s effort, in collaboration with the Centre for the Fourth Industrial Revolution India, to develop a roadmap for establishing an AI sandbox ecosystem tailored to India’s unique needs and sectoral priorities.40 Advancing Responsible AI Innovation: A Playbook 17