According to CNBC, the White House is finalizing a voluntary framework with OpenAI, Anthropic, and Google that would shift control over who can access the most powerful AI models from the companies to the government. Reported July 17, the plan, first surfaced by the Financial Times earlier this month, would give federal agencies up to a 30-day security review before new frontier models ship, and set benchmarks for release timing and access inside and outside the U.S. It stops short of a formal licensing regime. An announcement could come as soon as early August, when a deadline from a June executive order expires. It marks the most significant U.S. move yet to oversee frontier AI. What to watch: whether standards labeled voluntary start to feel mandatory.