🔍 Read the full analysis: What Happens Before Powerful AI Ships—and Who Pays For Testing? on ThorstenMeyerAI.com
Get hardware and tech essentials delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
The UK Budget confirmed an additional £120 million for the AI Security Institute across four financial years, supporting frontier AI evaluation and testing. The source reports that UK and US evaluations are voluntary and cites tests that found risky model behavior, but the funding comparison depends on time frame and US budget figures remain unsettled.
The UK Budget confirmed an extra £120 million for the AI Security Institute over four financial years, funding work to test frontier AI systems before or around release. The investment increases the institute’s stated existing annual funding of £66 million, while raising questions about how much testing can be required when evaluations in both the UK and US are voluntary, according to the source report.
The Budget allocates the new money across 2026-27 to 2029-30. Dividing £120 million evenly across those four years gives an average of about £30 million per year, although the Budget does not specify how the cash will be distributed year by year. AISI says it already receives £66 million per financial year, has more than 100 technical staff and priority access to more than £1.5 billion in public computing resources.
The new funding sits within a broader £700 million AI security package, as described by the Treasury. The source says the package also includes £165 million for a National AI Biosecurity Initiative and £40 million for a National Agentic Security Unit, intended to detect, investigate and respond to rogue agentic systems. The Budget describes AISI’s additional funding as intended to strengthen the UK’s frontier AI evaluation and testing capability.
The source compares UK funding with the US testing center at NIST, whose current name on its website is the Center for Advancing Innovation and Standards for Super Intelligence, or CAISSI. It reports about $15 million for the US center in fiscal 2026. The White House has requested $27 million for fiscal 2027, while a House committee report provides for up to $15 million. Those figures represent different stages of the US budget process; the requested and proposed amounts are not final enacted funding.
Infographic 01 · The money, made comparable
Britain spends about 8 to 9 times more per year. Not “10 times” from the top-up alone.
The new £120 million for the UK AI Security Institute covers four financial years. Put both countries on a per-year basis and the gap is still large, just smaller than the headline.
Testing budget per year
Million US dollars · UK at about $1.33 per pound, the rate Semafor used
Why “10 times” needs a footnote
- Semafor: the £120m (~$160m) top-up “alone is more than 10 times” the US center’s annual budget
- But: the £120m is spread over four financial years, the US figure is one year
- Per year: the top-up is about $40m, roughly 3 to 4 times the US budget
- With AISI’s existing money: about 8 to 9 times
- Not counted: free compute and model access from companies, and testing by other agencies such as the NSA or the UK’s NCSC
Institute for Progress estimates: a “limited” US center needs $26m a year, an “equipped” one $84m. Matching the UK’s share of total government spending would mean about $850m a year.
Sources: HM Treasury, Budget 2026, 28 Oct 2026; AISI About page, accessed 29 Oct 2026; Institute for Progress, 13 May 2026; Semafor, 27 Oct 2026; House CJS report FY2027 (up to $15m). Per-year averages, ratios and dollar conversions are my arithmetic.
Powered by Thorsten Meyer AI
Funding Before Frontier Models Ship
Government testing can give policymakers and developers information about how powerful models behave under risky conditions, including attempts to bypass safeguards or carry out cyber tasks. The source reports that UK evaluations this year found concerning behavior in simulated and live settings. Those findings are relevant to companies deploying AI agents, because systems able to take actions online can create risks beyond the production of text.
The funding increase does not, on its own, establish that every leading model will be tested or that a test will prevent harmful use. The source says evaluations by the UK institute and its US counterpart are voluntary. That limits how far public testing bodies can reach if a developer declines to share a system. For users, a government evaluation is evidence that investigators examined a model; it is not a guarantee that the model is safe in every setting.
For companies, public evaluation reports can offer practical examples of failure modes to check in their own systems. But testing arrangements, access to models and legal duties differ by jurisdiction. Businesses still need to identify which rules apply to their products and cannot treat a government test as a substitute for their own risk controls.
Products Worth Considering
Anker SOLIX C1000 Gen 2 Portable Power Station
Bose QuietComfort Wireless Noise-Cancelling Over-Ear Headphones, Black
DJI Osmo Pocket 3 Vlogging Camera with 1-Inch CMOS Sensor and 4K Video
As an affiliate, we earn on qualifying purchases.
How UK and US Budgets Compare
The headline comparison needs a consistent time frame. The source notes that describing the British top-up as more than ten times the US center’s annual budget compares four years of UK funding with one year of US funding. On an annualized basis, the £120 million top-up averages £30 million, and the source combines that with AISI’s existing £66 million annual funding to estimate roughly £96 million a year. Using the source’s approximate currency conversion, that is about $128 million annually, or around eight to nine times the reported $15 million US figure. Compared with the White House’s $27 million request for fiscal 2027, the UK estimate is under five times as large. These are approximate comparisons, not a measure of equivalent staffing or testing capacity.
AISI was established in 2023 and is part of the Department for Science, Innovation and Technology. The source says the 2025 Spending Review had already set aside £240 million for the institute. The US body also began in 2023 and has been renamed more than once; the source describes its role as voluntary evaluation focused on demonstrable risks such as cybersecurity and biosecurity. It reports that the center has no permanent director, following Chris Fall’s resignation in July, and that NIST does not publish a staff count for it.
The source also points to the EU AI Office as a different model. It reports that the office has more than 125 staff and, since August 2, 2026, has legal authority to evaluate large models and fine makers up to 3% of global turnover. That legal authority distinguishes the EU approach from the voluntary UK and US testing described in the report.
Testing Access and Funding Gaps
The Budget confirms the total UK top-up and its four-year period, but it does not specify how the money will be allocated by year or detail the staffing, evaluations or other capabilities it will fund. The source’s annual comparison uses an even-spending average; actual spending could differ.
US funding is also unsettled beyond the reported fiscal 2026 figure. The White House’s $27 million request and the House committee’s provision of up to $15 million are not the same as a final appropriation. The source does not give an official headcount for CAISSI.
The scope of access is another open issue. The source says testing is voluntary in both countries and reports that Anthropic released Mythos 5.1 in September without giving any organization outside the US access. It does not establish which organizations received access, what testing took place elsewhere, or whether the UK’s new funding will change developers’ participation.
How New Testing Capacity Develops
The next confirmed step is the start of the UK funding period in 2026-27, followed by spending through 2029-30. The Budget does not set out a year-by-year schedule or name particular models for evaluation, so further details from AISI or the government will be needed to show how the additional money is put to work.
In the US, the fiscal 2027 funding request remains subject to the federal budget process. The final amount will determine whether CAISSI receives the requested $27 million, the House committee’s proposed amount of up to $15 million, or another figure. The source also reports a bipartisan proposal for $100 million annually through 2029, but says that proposal is not law.
Readers and businesses can follow future public evaluation reports for evidence about what models were tested and under which conditions. Until access and participation rules change, the extent of testing will depend in part on developer cooperation. A test result should be read as a report on a model’s behavior under specified evaluations, not as a blanket approval for every use.
Key Questions
How much new funding did the UK announce for AI testing?
The UK Budget confirmed £120 million over four financial years, from 2026-27 through 2029-30, for the AI Security Institute’s frontier AI evaluation and testing capability.How does the UK funding compare with the US center’s budget?
The source reports about £96 million a year for the UK institute when its existing £66 million annual funding is combined with the new top-up’s £30 million annual average. It compares that with about $15 million for the US center in fiscal 2026, while noting the exchange rate and differences in budgets. The White House has requested $27 million for fiscal 2027, which is not confirmed final funding.Can governments require frontier AI companies to submit models for testing?
The source says evaluations in the UK and US are voluntary. It describes the EU AI Office as having legal authority to evaluate large models and impose fines under the rules that took effect on August 2, 2026.Does a government AI test mean a model is safe?
No. A test examines a model under particular conditions and may identify risks, but it cannot guarantee safe behavior in every use. The source cautions that government testing should not be treated as a seal of approval.Source: ThorstenMeyerAI.com
