The United States’ AI testing agency has a budget of about $10 million to $15 million a year and no permanent director, while the White House is asking Congress to raise its funding to $27 million. A White House spokesperson told Semafor the request, included in President Donald Trump’s FY2027 budget proposal, would be a 145% increase over the agency’s current budget.
The agency, now called the Center for Advancing Innovation and Standards for Super Intelligence, or CAISSI, sits within the Commerce Department’s National Institute for Science and Technology. It works with leading AI labs to assess models before release and calls itself the US government’s “industry’s primary point of contact.” Those arrangements are voluntary: companies are not legally required to submit models for review.
Why is the agency under scrutiny?
CAISSI was formed under the Biden administration in the same year as its British counterpart, the UK AI Security Institute. The US agency was originally named the AI Safety Institute. In 2025, the Trump administration removed “safety” from its name, reflecting its stated preference for accelerating AI development. Its previous director, Chris Fall, resigned in July, and the agency has not appointed a permanent replacement.
Arthur Tellis, a senior technology fellow at the Washington think tank Institute for Progress, told Semafor that the UK institute is impressive partly because Britain has invested heavily in it. Tellis argued that more money alone would not be enough for CAISSI: he said it also needs technically capable leadership, a direct line to the White House and a structure that lets it follow and influence the development of leading AI systems. He described the US agency as too small and too distant from decision-makers to do that effectively.
Congress has proposed a larger budget
Some lawmakers and White House science director Michael Kratsios have called for more resources. Representatives Lori Trahan, a Democrat from Massachusetts, and Jay Obernolte, a Republican from California, proposed a bill earlier in 2026 that would give CAISSI $100 million annually through 2029. Trahan said the agency needs funding to recruit skilled staff, support durable safeguards and help the US shape international standards.
The White House’s $27 million request and the lawmakers’ $100 million proposal are different funding plans, not approved budgets. The supplied details do not establish whether either will pass. Even with more money, CAISSI’s work with AI companies would remain voluntary unless the law changes.
Why it matters
For people who use AI products, CAISSI’s reviews could inform how the US government understands new models before they reach the public. But the agency’s current reviews are voluntary, and neither proposed funding increase is confirmed, so the details do not show that companies will face new testing requirements.
