Experts Call for New AI Standards Agency to Set Rules for Fairness and Safety
-
Licensing AI models is impractical due to the proliferation of open-source models and would be anti-competitive. However, some licensing ideas have merit, like establishing standards and a regulatory agency.
-
There must be technical and behavioral standards for developing and operating all AI systems to address negative effects, not just large proprietary models.
-
An outcomes-focused regulatory approach is needed that focuses on preventing discrimination, fraud, privacy harms, and other issues, regardless of the AI system.
-
Multistakeholder processes involving government, industry, and civil society should develop behavioral standards for AI's effects on individuals and society.
-
A new expert federal agency is needed to oversee standard-setting processes and enforce implementation of behavioral standards for all AI systems.