The Trump administration's AI regulatory framework is nearing finalization, with a draft already sent to OpenAI, Anthropic, and Google.

```

The Trump administration is accelerating the development of a regulatory framework for cutting-edge AI models, requiring companies to submit their most advanced models for government review before public release. The relevant draft has been circulated to major tech giants several days before the deadline.

On Tuesday, according to The Information citing three people familiar with the matter, the White House Office of the National Cyber Director distributed the draft framework to OpenAI, Anthropic, and Google about two weeks ago. The three companies then jointly submitted revisions. The creation of the framework stems from an executive order signed by Trump in early June this year, setting a completion deadline of August 1. The framework aims to establish a more standardized process for government review of powerful AI models, with a focus on frontier models capable of quickly discovering cybersecurity vulnerabilities.

This move will provide clearer regulatory rules for the previously uncertain AI regulatory environment. Since the executive order was signed, the White House has imposed export controls on Anthropic and required OpenAI to release its latest model GPT-5.6 in phases. These actions have triggered industry criticism that the government is imposing a temporary permit system on the AI sector. The introduction of this framework is seen as an important move by the White House to promote the development of the U.S. AI industry amid fierce competition with China.

Origin of the Framework: Executive Order to Address Safety Risks of Frontier Models

The direct trigger for the executive order was Anthropic’s frontier model, Mythos. Because this model possessed offensive cybersecurity capabilities, Anthropic voluntarily postponed its large-scale public release. While seeking countermeasures, the Trump administration strived to maintain its “hands-off” regulatory stance on AI while ensuring effective oversight of national security risks posed by new models.

According to sources familiar with the matter, the final version of the executive order underwent multiple rounds of negotiation. Former head of AI and cryptocurrency affairs David Sacks intervened at the last moment, blocking Trump from signing the initial order on the grounds that some clauses were too harsh for companies. Later, Trump signed a revised version on June 2, requiring AI labs to provide the federal government with up to 30 days of access before releasing “covered frontier models” to other partners, and giving all government agencies 60 days to design a concrete framework.

Since June, the White House has held multiple rounds of feedback meetings with companies, including a joint meeting on June 9, to complete the framework ahead of the August 1 deadline.

Core Dispute: How to Define Frontier Models

One of the key points of disagreement in the framework lies in how to define the scope of “frontier models.” According to two sources familiar with the matter, some smaller AI firms are concerned that the framework’s definition is mainly referenced to the three biggest labs—OpenAI, Anthropic, and Google—possibly excluding them.

Criteria for determining frontier models span multiple dimensions, including model size, deployment mode, and specific application capabilities as measured by third-party evaluation. If the definition focuses only on leading closed-source models like Mythos and GPT-5.6, it may exclude open-source models—which are typically trained with fewer parameters and often deployed locally rather than in the cloud—even if their capabilities are comparable to closed-source models.

For companies included in the framework, there are potential benefits: the executive order will direct more funding towards critical infrastructure and cybersecurity defense, although the size and actual allocation of these resources remain unclear.

Open-Source Models: Disputes over Regulatory Exemption

The treatment of open-source models is another unresolved core issue. According to a previous report by The Information, companies and the government have had parallel discussions about whether open-source models should receive discretionary review exemptions based on their capabilities, so as to reduce extra regulatory burdens.

Voices supporting the exemption argue that open-source models publicly available in the United States still lag behind their Chinese counterparts, with figures such as David Sacks advocating for reduced regulation to ensure room for catch-up. If the framework does not classify open-source models as frontier models, these models will not be able to participate in the program and will be cut off from related resources.

The Actual Binding Force of the Voluntary Framework Remains Uncertain

Although the Trump administration insists that compliance with the framework is not mandatory, its actual binding force remains one of the industry’s major concerns. The White House’s previous decision to impose export controls on Anthropic has led to doubts about the credibility of “voluntary” commitments.

At a news conference in May this year, when a senior White House official was asked about the consequences if companies refused to submit their models, the official did not respond directly but stated that the framework “aims to flexibly adapt and advance in step with new technology developments,” and said it “strikes an appropriate balance between safety and innovation.”

In terms of the review mechanism, the White House has told some companies that agencies responsible for implementing model reviews will be the National Security Agency (NSA) and the AI Standards and Innovation Center (CAISI) under the Department of Commerce. According to people familiar with the matter, CAISI came under political pressure during the Trump administration, but there is now growing optimism about the White House recognizing its technical expertise.

Risk Warning and DisclaimerThe market involves risks, and investment should be approached cautiously. This article does not constitute personal investment advice, nor does it consider the specific investment objectives, financial circumstances, or needs of individual users. Users should determine whether any opinions, views, or conclusions in this article suit their particular situation. Investment based on this article is at your own risk. ```