Towards AI Transparency and Accountability: A Global Framework for Exchanging Information on AI Systems

A global standard for AI information exchange can accommodate different regional regulations: The paper proposes that rather than creating conflicting regulatory frameworks worldwide, countries should adopt a common format for sharing AI system information that allows local regulations (like the EU AI Act) to coexist without incompatibility.

"AI cards" could make AI systems more comparable to consumers: Similar to nutrition labels on food, the proposed "AI cards" would display standardized, transparent information about AI systems' performance and risks, enabling the public to meaningfully compare competing AI products and make informed choices.

Automated assessments and benchmarks can reduce regulatory burden: The framework suggests using automated tools to evaluate AI systems against standardized measures, which could make compliance more cost-efficient and scalable for companies while reducing the administrative burden on regulators.

Collaboration between regulators and industry is essential for effective oversight: Rather than imposing regulations unilaterally, the paper emphasizes that regulators and AI developers should work together to create solutions that are both protective and practical, preventing "overregulation" that could stifle beneficial innovation.

High-risk AI applications require tailored transparency measures: The framework proposes that transparency requirements should be customized based on the specific risks posed by different AI applications, rather than applying one-size-fits-all standards across all AI systems.