US AI safety agency CAISI head resigns after 3 months | as Chinese open-source models gain US ground
TL;DR
Chris Fall, head of the US Center for AI Standards and Innovation (CAISI), resigned three months into the role. NIST director Arvind Rajaraman will serve as acting head. The timing is sensitive — Chinese open-source models are gaining traction in the US, competing directly with OpenAI and Anthropic.
The head of the US Center for AI Standards and Innovation (CAISI), Chris Fall, resigned three months after being tapped for the role by the Trump administration. CAISI sits under the Commerce Department and was set up to help the government promote "testing and collaborative research" around commercial AI systems. Commerce spokesperson Kristen Eichamer said NIST director Arvind Rajaraman will serve as acting head and continue to oversee the agency. Eichamer did not disclose why Fall stepped down.
Fall was picked to run CAISI in April. CAISI's predecessor was the US AI Safety Institute (USAISI), stood up under Biden; the Trump administration rebranded it and narrowed the mandate — from "model safety red-teaming" toward "industry standards coordination."
The timing is uniquely bad for the White House — Chinese open-source models (DeepSeek, Kimi, Qwen, GLM) are gaining ground in the US market, competing head-on with OpenAI and Anthropic on closed models. CAISI's evaluations are the pivot on whether Trump's team lists Chinese models as "untrusted" — a leadership gap here is a direct hit at the administrative layer.
Industry rumors say Fall was unhappy that CAISI got downsized into an "industry-dialogue tool," but Commerce declines to comment.
via CNBC
Fall was picked to run CAISI in April. CAISI's predecessor was the US AI Safety Institute (USAISI), stood up under Biden; the Trump administration rebranded it and narrowed the mandate — from "model safety red-teaming" toward "industry standards coordination."
The timing is uniquely bad for the White House — Chinese open-source models (DeepSeek, Kimi, Qwen, GLM) are gaining ground in the US market, competing head-on with OpenAI and Anthropic on closed models. CAISI's evaluations are the pivot on whether Trump's team lists Chinese models as "untrusted" — a leadership gap here is a direct hit at the administrative layer.
Industry rumors say Fall was unhappy that CAISI got downsized into an "industry-dialogue tool," but Commerce declines to comment.
via CNBC
