Oversight and approval of algorithms

Comprehensive frameworks on governance

Governance frameworks should: • Support increasing collective understanding of the AI phenomenon and collective reflection and informed policy decisions on the need for and means of governing AI • Facilitate consensus-building between various stakeholders and support cost–benefit analyses between values and interests • Enable meaningful stakeholder and public consultation and participation in decision-making • Use a coherent and integrated set of tools that combines various solutions, tools, and techniques at different levels of society to improve decision-making • Ensure that developers of AI systems are subject to statutory oversight by an independent regulator with appropriate investigative and enforcement powers • Account for the diversity of AI technology and services, including their short- and long-term direct and potential indirect impacts and challenges

Processes for improving public governance and coordination of AI

AI governance should implement agile and adaptive governance processes and pay attention to the following recommendations: • Stakeholder collaboration and public deliberation should be maintained as key inputs throughout governance processes. • Adopt responsible innovation principles and processes, including communicative principles for deliberation • Use co-regulation process in developing AI regulation • Coordination organizations can support agile and adaptive governance and co-regulation • Ensure that governance strategies are based on understanding the long-term consequences and challenges of AI governance

Ethics and human rights in policy making

Human-rights standards and approaches-based actionable ethical principles can be enhanced using: • Assessment of governance capacities and dynamics; ethical and human-rights risks • Collaboration and stakeholder participation • Operationalizable tools, mechanisms, and recommendations • Prerequisites for operationalization include oversight structures, accountability, traceability, sanction mechanisms, and design for supporting stakeholder involvement and value alignment in AI development and use

Tools and tasks for governing institutions

A regulatory agency or relevant governing institution to support operationalization of good governance and ethical principles. Tasks of such governing institutions include: • Oversight and approval of algorithms • Supervision of organizations developing AI • Assessment of ethical issues and social impacts of AI, data governance, risk-management mechanisms, impacts on and of legislation, and AI development processes • Certification, audition, and development • Testing and licensing

Agile and adaptive governance processes and coordination

• AI governance should utilize adaptive, people-centered, and inclusive policy-making, as governance is a result of multi-stakeholder action coordinated by the state • It should adopt decentralized, bottom-up decision-making, drawing on an array of expertise within and outside public administration, and broad participation • In addition to ethical principles, lessons should be learned from the RRI approach on how to systematically address societal challenges in technology development and use • Adoption of transformative innovation policy—innovation policy is also about tackling societal challenges • Communicative principles of deliberation and RRI should be used in AI governance and policy-making • Establishment of governance coordinating committee or similar organization to coordinate, e.g., AI stakeholder engagement, dialogue, recommendations, and guidelines

Co-regulation processes in developing AI regulation

Develop regulation according to a co-regulatory model, where industry and other stakeholder representatives together with the public administration negotiate statutory obligations