Chris Fall has stepped down from his role as the director of the Center for AI Standards and Innovation (CAISI), an announcement confirmed by the agency through various news sources. Fall's tenure lasted only three months, following the brief appointment of Collin Burns, who resigned after less than a week. According to reports from The Washington Post, Burns was reportedly "pushed out" in April due to his prior affiliation with the AI company Anthropic, which had faced tensions with the Trump administration.
The reasons behind Fall's exit remain unclear. Before taking the helm at CAISI, he had served as the director of the Department of Energy’s Office of Science during the first Trump administration and had also led the DOE’s Advanced Research Projects Agency-Energy. His background includes a position in the DOE’s Office of Naval Research.
The agency was previously managed by venture capitalist David Sacks, who acted as the White House AI and cryptocurrency czar until his departure in March.
CAISI functions under the auspices of the National Institute of Standards and Technology and is the main body responsible for establishing technical standards and evaluation methods for AI models, as well as identifying cybersecurity risks. Nonetheless, it was not the organization at the heart of a recent controversy regarding model risks.
This unrest surfaced in June when the U.S. Commerce Department activated a little-known export control regulation that compelled Anthropic to withdraw its Mythos and Fable AI models from the market. The ban was lifted later that month after Secretary of Commerce Howard Lutnick expressed satisfaction with Anthropic's safety measures.
Additionally, earlier this month, the White House enacted an executive order establishing an AI safety oversight initiative named “Gold Eagle.” This program aims to coordinate efforts related to cybersecurity vulnerabilities and includes several federal agencies, including the Commerce Department and Department of Homeland Security, although CAISI was conspicuously absent from the list.
Following the lifting of the ban on Anthropic’s models, Google DeepMind CEO Demis Hassabis advocated for the formation of an independent standards organization for frontier AI, similar to the Financial Industry Regulatory Authority (FINRA). This aligns with CAISI's original mission.
Fall’s departure coincides with recent concerns over the competitive capabilities of the Chinese AI lab Moonshot, which unveiled a new version of its open model, Kimi. This model has demonstrated strong performance against other leading AI systems, prompting discussions within the administration about possibly blocking access to Chinese open models. This issue sparked significant debate and outcry over the weekend, with Sacks arguing against using regulation as a means of protectionism for U.S. AI companies.
While CAISI has published a few analyses of the capabilities of Chinese open-weight models such as Z.ai’s GLM-5.2 and DeepSeek V4 Pro, it has remained relatively quiet about its testing methodologies. Open weight signifies that such models can be downloaded and executed locally, although their training code and datasets are not publicly available. Since July 9, TechCrunch has reached out to both the Department of Commerce and NIST multiple times to inquire about the LLM evaluation process, but has yet to receive a response.
