US AI safety center staff asked to report informally to White House office
For a model-risk officer at a regional bank
Staff at the Center for Advancing Innovation and Standards for Super Intelligence (CAISSI), a Commerce Department division, were recently asked to informally report to an office of the White House, according to two former government officials who spoke to The Associated Press anonymously.1
The sources do not say why the White House asked for the arrangement; the White House did not respond to specific questions about CAISSI, the AP reported.1 The request is secondhand: it rests on two anonymous former officials, and the AP did not name the White House office or report any confirmation from the administration.1
CAISSI, until recently called the U.S. Center for AI Standards and Innovation, is allocated $10 million per year, about one-eighth the annual budget of its counterpart in the United Kingdom, the AP reported.1
The division's director was hired and then fired within days in April, and his replacement resigned after a few months, according to the AP; the head of its parent institute is acting director.1
In August, Congress started asking why the center stopped making public its reports on how safely certain AI systems performed, the AP reported.1
In September, Trump asked leading AI companies to undertake "tremendous self-policing" under a voluntary safety pact, and his administration has not said how it will monitor implementation, according to the AP.1
Banks buying or building on frontier AI models cannot count on published federal test results, so model risk and vendor due diligence will have to rely on the vendor's own evaluations, outside evaluators or the bank's own testing. Agentic systems are the hardest case: Sayash Kapoor of UC Berkeley told the AP no standard process exists, and checking only the final result can miss serious errors in the steps.
Our take. The government's AI safety center gets $10 million a year, has no permanent leader and stopped publishing its tests. When examiners ask who vetted your frontier model, the answer has to be your own team.
Action. Model risk: Document in-house validation for each frontier model and AI agent, testing every step the agent takes, not only final outputs.
What our records say
Commerce's AI strategy is an AI pillar within its enterprise data strategy plus a written roadmap, with AI-driven receivables automation for commercial clients as the main client-facing use.1 Commerce discloses an AI pillar within its enterprise data strategy and a written AI roadmap, and for commercial clients it provides AI-driven receivables automation, with clients such as AlloSource (on the Esker platform) and Interstates using automated collections and cash application.1
From our records
- White Houseregulator · binding Non-binding standards · latest EO 14365 (Dec 2025), America's AI Action Plan (Jul 2025)
- Commerce Bancsharesbank · latest Nolan & Associates acquisition completed (Oct 2026)
- Associated Banc-Corpbank · latest Barclays conference: 30 AI use cases in process (Sep 2026) · vendors NICE Actimize