According to a report by Wired magazine citing sources, the current U.S. cybersecurity review mechanism for cutting-edge artificial intelligence models is "almost certainly" expected to expand in the coming months to open-source (open-weight) models. The term "cutting-edge" refers to models with capabilities comparable to Anthropic's Claude Mythos or OpenAI's GPT-5.6, but the report does not provide more detailed specifications of corresponding versions.

Insiders revealed that the current review framework remains voluntary. Some U.S. officials have realized that mandatory reviews might stifle AI development, and therefore are temporarily reluctant to make it a mandatory compliance requirement. However, the signal that the scope of review is expanding from closed-source to open-source models is very clear, indicating that companies such as DeepSeek and Meta, which focus on open-source approaches, may also face U.S. government cybersecurity scrutiny in the future.

Open-source models face a 'market squeeze' effect

The ripple effects of the expanded review go beyond compliance itself. Once closed-source models pass the review and receive official endorsement, U.S. companies may lose interest in using unreviewed open-source models, which could discourage local vendors from developing open-source models. This is one of the factors considered by the White House. In other words, even if the review is technically voluntary, the 'soft constraints' in the market are already sufficient to compress the space for open-source models.

At the same time, there has long been a call within the U.S. government to tighten regulatory oversight, pushing for a gradual transition from voluntary reviews to mandatory requirements. The core issue surrounding open-source models lies in the fact that open weights allow anyone to download and deploy the model, making security risks harder to control. However, over-tightening regulations could stifle the innovation vitality of the open-source ecosystem and push developers to overseas markets with less stringent regulations. Finding a balance between security assurance and open innovation will be a key focus of U.S. AI policy in the coming months. This decision's impact will extend far beyond the United States, directly reshaping the global competition between open-source and closed-source large models.