Khanna Raises Alarm Over Center for AI Standards and Innovation’s (CAISI) Lack of Permanent Leadership, Gaps in Assessing PRC AI Model Safety

WASHINGTON – Representative Ro Khanna (CA-17), Ranking Member of the House Select Committee on the Strategic Competition Between the United States and the Chinese Communist Party (CCP) sent a letter to Secretary of Commerce Howard Lutnick seeking information regarding the Center for AI Standards and Innovation’s (CAISI) lack of permanent leadership and recent cessation of publishing information on People’s Republic of China (PRC) AI models’ propaganda and censorship.

Khanna cites reporting that in the absence of permanent leadership, CAISI has ceased publishing information on content manipulation in its latest evaluations of PRC AI models leaving critical gaps in assessing the safety of increasingly advanced PRC AI. Khanna warns that the lack of a permanent director and resulting unstable environment threatens CAISI’s directive to help identify, mitigate, and prevent AI-facilitated disinformation and misinformation.

“CAISI cannot be without a leader amid this critical moment in PRC model development,” Khanna warns. “While the PRC’s advances would pose challenges for the U.S. Government under any circumstances, the situation is undoubtably far riskier due to CAISI’s lack of a permanent director and its refusal to publish findings on PRC AI content manipulation.” 

Khanna requests answers to the following questions by September 1, 2026:

• Public reporting indicates that (a) the White House abruptly ousted both Collin Burns and Chris Falls from their leadership posts, (b) CAISI removed documentation regarding model-testing agreements with tech firms, and (c) CAISI was blocked from publishing model assessment reports. What were the specific justifications for each of these actions?
• CAISI and UK AISI’s preliminary evaluation found that Moonshot’s Kimi K3 model’s guardrails do “not prevent it from attempting cyber exploit development or offensive cyber operations.” Security experts have also expressed concern that during third-party performance testing, Moonshot’s Kimi K3 AI model escaped its sandbox environment. How do you respond to these security vulnerabilities? And how, if at all, does the Department plan to ensure that future PRC AI model releases do not cause similar cyber-incidents? 
• Given that some experts believe models like Kimi K3 indicate that the PRC is only 3-6 months behind the U.S. in frontier AI, how does the Department plan to measure and mitigate the national security risks posed by advanced PRC open-weight models while CAISI lacks a permanent director? 
• Will CAISI test future PRC AI models for their ability to generate disinformation, misinformation, propaganda, or any other forms or methods for generating or aligning with Chinese Communist Party talking points and censorship? 

Read the full letter here.