The article exposes the revolving door of leadership at the US government’s AI standards body, CAISI, with director Chris Fall resigning after just three months. This follows the departure of previous appointee Collin Burns in less than a week, reportedly due to his prior work at Anthropic, which the Trump administration was battling. Venture capitalist David Sacks, who had earlier held the White House AI and crypto czar role, also stepped down in March. The agency, tasked with developing technical standards, testing methods, and cybersecurity risk assessment for AI models, has been unable to maintain stable leadership, raising questions about the government’s ability to shape AI regulation. The article notes that TechCrunch has sent multiple inquiries to the Department of Commerce and NIST about CAISI‘s LLM evaluation processes and has received no response.

The instability is compounded by concrete regulatory actions and debates. The Commerce Department used an obscure export control directive to force Anthropic to pull its Mythos and Fable models, though the ban was later lifted after Secretary Howard Lutnick accepted Anthropic‘s safety plans. Meanwhile, the White House’s “Gold Eagle“ executive order for AI safety oversight notably excluded CAISI from the list of participating federal organizations. This void has led figures like Google DeepMind CEO Demis Hassabis to call for an independent, industry-run standards body modeled after FINRA. Separately, the administration has been weighing efforts to ban Chinese open-weight models, sparking debate from Sacks, who argued regulations shouldn’t be used as protectionism. CAISI has released reports on capabilities of Chinese open-weight models like Z.ai’s GLM-5.2 and DeepSeek V4 Pro, but has not been transparent about its own testing processes.

For builders, the takeaway is that the US regulatory landscape for AI remains fragmented and politicized. The rapid turnover at CAISI and its exclusion from the Gold Eagle program suggest the government lacks a coherent strategy for technical standards. Serious engineers should prepare for continued uncertainty, watch for industry-led standards initiatives, and note that debates over open-weight models (like Moonshot’s Kimi) could lead to sudden market restrictions. The lack of transparency around CAISI‘s evaluation methods also means that relying on government certifications for model safety is premature. The article ultimately shows that the organizations meant to guide AI standards are themselves unstable and sidelined, leaving the private sector to fill the gap.