OpenAI confirmed its agreement to comply with President Donald Trump's executive order requiring AI companies to allow federal government assessment of AI models before public release. The company will submit models 30 days prior to launch for government benchmarking to evaluate advanced cyber capabilities and determine if they qualify as “covered frontier models” [1, 2].
President Trump signed the executive order on June 2, 2026, mandating government review of advanced AI models before they become publicly accessible [1]. Initially, the order required a 90-day prior submission period, but after feedback from the tech industry, the administration shortened that window to 30 days [2]. Some sources note this change while others only reference the 30-day requirement without the original 90-day rule [1, 2].
OpenAI’s head of countries, George Osborne, expressed support for government oversight of AI development. He suggested that regulatory bodies should have the authority to evaluate AI systems, but also require enough flexibility to adapt to advances in the technology [1, 2].
The order remains voluntary, requesting companies’ participation in benchmarking but not imposing strict requirements. Critics argue it falls short of adequately regulating potentially dangerous AI models, calling the policy underwhelming given the risks posed by advanced AI systems [1, 2].
OpenAI announced its compliance with the executive order on June 5, 2026, signaling cooperation with the government’s efforts to assess and manage the risks associated with frontier AI technologies [1, 2].
