Effectiveness of generative AI regulation methods

Determine which regulatory methods for generative artificial intelligence are effective in practice, identifying policy instruments that reliably achieve intended outcomes for systems such as ChatGPT.

Background

The paper investigates one specific regulatory approach—geographic restrictions on access to ChatGPT—and finds that such restrictions did not reduce use in scientific preprints and were often associated with higher usage in restricted countries.

Despite this domain-specific evidence, the broader question of which regulatory methods work across contexts remains unresolved, highlighting a need for rigorous, empirical evaluation of policy instruments for generative AI.