Aaron Levie discusses current implicit AI regulation
According to a post on Marginal Revolution, Box CEO Aaron Levie stated that the United States now has de facto AI regulation, meaning models with certain capability levels or trained on specific compute sizes may require government review before release. Levie noted that as AI models became more powerful, such regulation was inevitable, though he believes it is too early. He outlined several implications, including that the U.S. can control who gets access to frontier intelligence as long as it remains at the frontier. Levie also suggested this could create a backlog of AI releases, reducing rapid model progress, and that other countries may have more incentive to pursue sovereign AI strategies. He identified open-weight models as a likely winner, as they can still be released without the same controls, and raised the question of how regulation might eventually extend to open models. Levie expressed satisfaction that significant early AI progress occurred before these controls were implemented.
What’s reported
Open questions
Key figures
Sources: marginalrevolution.com
