Staff changes are repeated by the AI Standards Agency of the National Institute of Standards and Technology (NIST). Several media sources cited confirmation that Chris Fall, head of the CAISI, had left his post about three months ago.

This is the second rapid replacement of the post in the recent past. The former successor Collin Burns left in less than a week. The Washington Post stated at the time that Burns had been forced out of office in a context of tension between the Trump Government and the company because of his employment in Anthropic. The reasons for Fall's departure are not yet known.

CAISI is responsible for model standards and safety tests

CAISI is part of NIST and is primarily responsible for the technical standards for the AI model, the test methodology and the safety risk assessment of the network. By function, it should have been one of the key institutions in the United States federal AI security system.

However, CAISI was not included on the list of participating agencies in the new AI Security Oversight Project “Gold Eagle”, launched by the White House this month. The project aims to establish a coordination mechanism for cyber-security gaps involving federal agencies such as the United States Department of Commerce and the Department of Homeland Security.

The Anthropic incident exacerbated regulatory differences.

U.S. AI regulatory controversy rose significantly in June. At that time, the United States Department of Commerce invoked an unusual export control directive, effectively forcing Anthony to withdraw the Mythos and Fable models from the market. By the end of June, the restrictions were lifted after the Minister of Commerce, Howard Lutnick, confirmed the security programme for Anthropic.

This incident has also drawn attention to the boundaries of responsibility between federal agencies. Although CAISI is responsible for standards and testing, the actual controversy surrounding the risk of front-line models in recent rounds has not been handled by the agency.

Industry for Independent Standards

Following the restoration of the Anthropic model, Google DeepMind Chief Executive Officer Demis Hassabis publicly called for the establishment of an industry-led, independent, front-line AI standard body, modelled on the Financial Regulatory Authority of the United States, FinRA. This idea overlaps with some of the tasks undertaken when CAISI was created.

At the same time, there are discussions within the United States Government on how to address the competitive pressures of China ' s open source model. Axios reported that the Government had weighed the promotion of a restricted open source model in China. Last weekend, discussions rose rapidly after the release of the new open source model Kimi by the China AI Laboratory Moonshot.

Some people, including former White House AI and encryption chief David Sacks, have publicly opposed the use of regulatory tools to protect United States closed-source AI. According to Sachs, regulation should not evolve into a means of industrial protection.

The testing process still lacks a public description

It was mentioned that CAISI had previously published a capability report on the open source weight model in China, covering the GLM-5.2 and DeepSeek V4 Pro of Z.ai, but that the public description of its own testing process remained limited. TechCrunch states that the United States Department of Commerce and NIST have been asked several times about the large model assessment methodology since 9 July, but have not yet received a response.

The ownership of the United States AI security standards system is still in the process of being adjusted as the leadership changes again, new White House projects are launched, and industry calls for independent standards bodies are rising.