The paper seeks to guide AI policymaking and regulation to ensure the development and deployment of accountable, responsible, and trustworthy AI. To that end, CIPL recommends a risk-based approach to AI regulation that builds on existing laws and standards, such as privacy, consumer protection, intellectual property and anti-discrimination laws, as well as on accountable AI governance practices of organizations. The ten recommendations discussed in the paper are as follows:
A. Principle-and-Outcome-Based Rules
- Create a flexible and adaptable framework that defines the outcomes to be achieved, rather than prescribing details of how to achieve them;
- Adopt a risk-based approach that considers risks and benefits holistically;
- Build on existing hard and soft law foundations; and
- Empower individuals through transparency, explainability, and mechanisms for redress.
- Make demonstrable organizational accountability a central element of AI regulations;
- Advance adoption of accountable AI governance practices; and
- Apportion liability carefully, with a focus on the party most closely associated with generating harm.
- Create mechanisms for coordination and cooperation across regulatory bodies;
- Institute cooperation-based regulatory oversight and enable ongoing regulatory innovation; and
- Strive for global interoperability.