Regulators Have Tools to Evaluate AI Systems, But Tailored Oversight Is Needed

Governments are implementing policies to regulate AI systems and discover harms, but regulators should also evaluate systems under existing laws.

The regulatory toolbox includes transparency requirements, audits, sandboxes, industry engagement, and whistleblowers.

Transparency can inform the public but requires tailored rules. Audits directly evaluate systems but need expertise.

Sandboxes facilitate communication but demand capacity. The AI assurance industry aids monitoring.

Complaints from individuals and whistleblowers provide unique insights into algorithmic systems.