Sources: Trump administration officials have told CAISI to halt publication of its model assessments while an EO President Trump signed last week is implemented
Unit's future was thrown into doubt after its public work was halted despite winning praise from AI developers
Wall Street JournalAmrith Ramkumar
Context & Ripple Effects
The administration’s approach to AI model oversight has been unsettled: earlier coverage described internal disputes over pre-release model access and over whether CAISI or another federal office should lead evaluations. The recently signed executive order appears to formalize a government-assessment role that OpenAI has said it will support.
Pausing CAISI’s public assessments puts a previously visible part of that oversight effort on hold while the new process is implemented, making the unit’s institutional role a central question rather than simply an operational one.
First-order effects
CAISI will stop publishing model assessments for now, removing a public output of the federal evaluation effort during the executive order’s rollout.
AI developers seeking clarity on federal evaluations face an interim process centered on implementing the order rather than CAISI’s published assessments.
Second-order effects
The pause may intensify the existing contest among administration offices over who controls model-evaluation policy and access to advanced systems.
Companies that had been preparing to accommodate government capability assessments may need to adapt to changing procedures, points of contact, and disclosure expectations.
Third-order effects
If implementation consolidates evaluation authority outside CAISI or limits publication, federal AI oversight could shift toward less transparent, executive-branch-led access to models rather than publicly visible assessments.
The episode underscores that pre-release model evaluation is becoming a durable policy lever, though its eventual institutional home and degree of public accountability remain unresolved.
The trend: This is part of a broader shift from debating whether the U.S. government should evaluate frontier AI systems to defining which agency gets access, authority, and control over that process.
Few things are more important than getting the White House and CAISI to work together better. No last-min leadership changes, no deleted blog posts... CAISI looms large today and even larger longer-term in any serious federal plan - need to get it right https://x.com/...
Halting CAISI's public evaluations of models is a huge mistake. CAISI's public eval of DeepSeek v4's capabilities and censorship was tremendous, and essential to pushing back on Chinese propaganda that Chinese models are superior options to US models. CAISI is a strategic asset! …
CAISI has reportedly been directed to stop publishing public model assessments as the new AI EO gets implemented. Natsec engagement on AI is essential. But pulling CAISI's evals from public view doesn't make the field more secure. It just means fewer eyes on the science when we […
The de facto lockdown of CAISI is extremely disheartening. Wish more AI industry leaders would speak out, and not merely through policy documents but to POTUS directly. @sama @elonmusk @demishassabis https://www.wsj.com/...