White House expected to expand AI oversight to open models
The White House is preparing to close the biggest gap in its new AI oversight framework: open-weight models. According to people familiar with the matter, the administration is almost certain to revise its AI guidelines and extend oversight to open models once they reach frontier capability.
The White House is expected to expand its AI oversight framework in the coming months to cover open models once they reach frontier capabilities, according to people familiar with the matter reported by Wired. Officials are "almost certain" to revise the administration's new AI guidelines, which currently center on a voluntary frontier-model review process that the government has been rolling out with OpenAI, Anthropic, Microsoft and other leading labs. Under that framework, a developer can hand a model to the government for up to 30 days before release — but it only covers closed models, a gap that has drawn sharp criticism from smaller labs and open-weight advocates since the framework's details surfaced last week. Fortune reported that the White House has kept the evaluation framework itself private, and that some in the industry fear open models' exclusion could push customers toward "government-approved" closed alternatives.
The timing is not subtle. Hours before Wired's report, Alibaba's Qwen team open-sourced Qwen3.8-2.4T-A95B, its first Max-class model — a frontier-scale system anyone can download and self-host, the exact category the expanded framework would now reach. We covered that release this morning — Qwen open-sources Qwen3.8-2.4T-A95B, its first Max-class model. The report lands as the White House also wrestles with what to do about Chinese AI, where open-weight releases from labs like Kimi and DeepSeek are the primary vector of frontier capability entering the wild.
The stakes go beyond process. A voluntary 30-day review is awkward for a closed model; for an open one it is nearly unenforceable — once weights are public, pre-release review is a courtesy, not a control. That is precisely why this expansion, if it happens, will be the framework's real test. Wired frames the White House as still grappling with "how to regulate a technology it has tried not to regulate," and open models are the hardest version of that problem: the line between "frontier capability" and everything else will decide which downloadable models get caught, and whether the US answer to open-weight competition is oversight or something closer to restriction.
What to watch: whether the expanded framework keeps review voluntary, and how "frontier capability" gets defined — that threshold will determine which open models fall inside the net.
Open-weight frontier models can be downloaded by anyone — can a pre-release review framework meaningfully govern them at all? Tell us in the comments.
Sources: Wired · Techmeme · Fortune — White House won't release the evaluation framework · Fortune — smaller AI labs push back · Qwen3.8-2.4T-A95B (Hugging Face)
