The UK government will add £120 million (about $160 million) to the UK AI Security Institute (AISI), a government laboratory that tests some of the world’s most powerful AI models. Chancellor of the Exchequer John Healey previewed the funding as part of a new AI package ahead of his budget presentation to Parliament on Wednesday.
On its own, the £120 million top-up is more than ten times the annual budget of the comparable U.S. entity. The funding forms part of a broader UK package aimed at bolstering security and research capacities amid rising concerns over AI capabilities and cyber risks.
Components of the UK package
Other elements of the package include:
- £60 million to a sovereign AI fund intended to support domestic AI companies;
- £18 million to bolster domestic chip design;
- £10 million per year for research into how AI is reshaping employment.
A senior UK government spokesperson told Semafor that Britain is well positioned to succeed, citing a trillion-dollar tech sector outside the United States, AISI’s role as a global leader in frontier model testing, and the depth of capital markets to scale companies.
About the AISI
Formed in 2023, the UK AI Security Institute has more than 100 staff, including alumni from labs such as OpenAI and Google DeepMind, as well as personnel from British intelligence agencies and universities. Healey described the package as “a distinctively British approach to AI,” and said it was unlikely to encounter trouble in Parliament where his Labour Party holds a large majority.
The U.S. counterpart and its constraints
In the United States, the Center for Advancing Innovation and Standards for Super Intelligence (CAISSI) — renamed in 2025 to reflect President Donald Trump’s preferred terminology — operates within the Commerce Department’s National Institute of Standards and Technology (NIST) on a budget of roughly $10–$15 million per year. CAISSI was established the same year as the UK institute during the Biden administration but has seen major changes since Trump returned to office.
CAISSI currently lacks a permanent director; its last director, Chris Fall, resigned in July. The Trump administration removed the word “safety” from the organization’s original name in 2025 to signal a tilt toward embracing AI acceleration.
Arthur Tellis, a senior technology fellow at the Institute for Progress in Washington, told Semafor that AISI is “an extremely impressive organization” largely because of the UK’s substantial investment. Tellis added that beyond funding, CAISSI needs technically competent leadership, a direct line to the White House, and an operational model that lets it track and shape frontier AI development.
Calls for higher US funding and legislative proposals
Some members of Congress and officials have pushed for more resources for CAISSI. A White House spokesperson told Semafor that President Trump requested $27 million for CAISSI in his FY2027 budget proposal, a 145% increase over the current funding level.
Representative Lori Trahan (D-Mass.) told Semafor that allies and adversaries recognize the need to build expertise to oversee frontier AI and warned the United States cannot fall behind. Earlier this year Trahan and Representative Jay Obernolte (R-Calif.) proposed a bipartisan bill that would allocate $100 million annually to CAISSI through 2029.
Other U.S. AI funding and initiatives
Some argue the U.S. is funding AI through other channels even if not concentrated on safety-focused institutes. The Department of Energy’s Genesis Mission recently announced $159 million in awards for AI-enabled quantum, science and energy projects.
President Trump has also sought to address safety concerns by convening AI executives at the White House to sign a nonbinding “Accord on Super Intelligence” and by establishing a new AI task force led by Director of National Intelligence Jay Clayton. A White House official told Semafor that American Super Intelligence models are the most powerful in the world and that CAISSI will continue to play a role in pre-deployment testing and in the broader Super Intelligence Force; senior CAISSI officials will collaborate with relevant White House colleagues.
Notes on reporting and cooperation
The New York Times ran a profile of the UK’s AI institute in May. Politico reported in September that the White House asked OpenAI and Anthropic to refrain from sharing new models with the UK agency until the U.S. government had the chance to review them.
The UK’s new funding commitments mark a substantial public investment in testing frontier AI models and supporting domestic AI capabilities, while U.S. efforts remain smaller in scale and spread across different agencies and initiatives.
