The White House has completed a framework for evaluating advanced artificial intelligence systems but reportedly intends to share it only with selected technology companies and other trusted partners.
By yourNEWS Media Newsroom
The Trump administration reportedly plans to keep its newly completed artificial intelligence testing framework out of public view after privately presenting the proposal to major technology companies.
White House officials met with AI industry partners Tuesday to discuss standards for evaluating the capabilities of certain advanced models before their public release. The administration does not intend to publish the framework, Axios reported, citing three people familiar with the plans.
The identities of the “trusted partners” that could receive access to frontier AI systems before those models are released to the public remain unclear.
President Donald Trump directed federal officials in a June executive order to develop an AI safety-testing framework. The resulting plan is expected to establish procedures for examining the capabilities of some of the most advanced systems developed by private companies.
The White House did not immediately respond to a request for comment from the Daily Caller News Foundation.
The administration’s approach will reportedly exempt open-source AI models from the initial testing requirements, according to a separate Axios report.
That exclusion would distinguish models whose underlying code or technical components are made broadly available from closed systems controlled entirely by their developers.
Open-source models may not remain exempt indefinitely. Nvidia and other companies developing such systems could eventually be required to submit them for government review as their capabilities become more advanced, The Wall Street Journal reported, citing people familiar with the matter.
Nvidia CEO Jensen Huang recently discussed AI oversight with White House officials. The company has also launched the Open Secure AI Alliance, an industry coalition that includes Palantir, IBM, CrowdStrike, SpaceX and other major technology businesses.
The alliance advocates for the continued development and availability of open-source AI systems while promoting security standards for the technology.
The reported exemption comes as the Trump administration appears increasingly focused on limiting access to advanced Chinese open-source models rather than imposing immediate testing requirements on comparable systems developed in the United States.
Chinese companies have narrowed the performance gap with American AI developers, increasing concern in Washington about the security, economic and technological consequences of widely available foreign models.
Moonshot AI, a Chinese laboratory, released its Kimi K3 model in July. The system has been described as a growing challenge to American leadership in artificial intelligence.
Trump administration officials have increased their scrutiny of Moonshot AI amid allegations that the Chinese company used Anthropic’s flagship model to help train Kimi K3 through a process known as distillation.
Distillation generally involves using the outputs or capabilities of a larger, more advanced system to improve a smaller or separate model. Administration officials reportedly believe Moonshot AI may have relied on Anthropic’s technology without authorization to accelerate development of Kimi K3.
The White House framework therefore emerges as the administration attempts to balance several competing objectives: strengthening oversight of powerful AI systems, protecting American technological leadership and avoiding restrictions that could slow domestic open-source development.
By withholding the full framework from the public, the administration would limit outside scrutiny of the standards used to assess advanced models and the process for selecting companies or organizations granted early access.
The policy’s final reach may depend on how quickly open-source systems advance and whether U.S. officials determine that models developed by companies such as Nvidia pose risks comparable to closed frontier systems.
For now, the framework reportedly places greater emphasis on private coordination between the White House and selected industry participants than on public disclosure or broadly applicable government review.
Original article
Comments