UK's AI Security Institute Publishes First Frontier AI Trends Report

The UK's AI Security Institute — known as AISI — has released its first Frontier AI Trends Report, drawing on two full years of internal frontier model testing.

At the same time, AISI published a study in the journal Science on AI persuasion — specifically, a large-scale experiment examining the mechanisms by which conversational AI influences people. The institute also announced a new research Memorandum of Understanding with Google DeepMind, expanding a collaboration that was already underway.

What AISI Actually Is Now

Most coverage frames AISI as a promising experiment. But AISI is now a functioning research organization inside the UK's Department for Science, Innovation and Technology, operating on £66 million per financial year in long-term committed funding. It has 100-plus technical staff. It has priority access to over £1.5 billion in compute through the UK's AI Research Resource and exascale supercomputing program.

The Leadership Roster Matters

Interim Director Adam Beaumont is the former Chief AI Officer of GCHQ — Britain's signals intelligence agency. Chief Technology Officer Jade Leung is simultaneously serving as the Prime Minister's AI Advisor and previously led the Governance team at OpenAI.

Chief Scientist Geoffrey Irving and Research Director Chris Summerfield have both led teams at OpenAI, Google DeepMind, and the University of Oxford.

Chair Ian Hogarth is a tech investor and entrepreneur. The advisory board includes Yoshua Bengio — one of the pioneers of modern deep learning and a vocal voice on AI risk.

What the U.S. Has By Comparison

When AISI was formalized, it had roughly 10 times the budget of the U.S. government's own AI Safety Institute, established at the same time. The United States invented this technology and American companies dominate the global AI market. Britain — with a fraction of the GDP — built the better-funded, more operationally advanced government safety testing apparatus.

What the Reports Contain

The Frontier AI Trends Report represents two years of testing data on the world's most capable AI models, now publicly available. The Science paper on AI persuasion is a peer-reviewed, large-scale experiment on how AI systems influence people's thinking — published in one of the most prestigious journals in the world.

Why This Matters

AI persuasion is not an abstract concern. These systems are already embedded in social media recommendation engines, customer service bots, and political advertising pipelines. A government body just published peer-reviewed evidence on how they work on human psychology.

The Frontier AI Trends Report is the first independent, state-backed public accounting of what today's most powerful AI systems can actually do — not what their creators say they can do. Independent testing of powerful technology is how markets function. It's how nuclear plants get audited. It's how drugs get approved. AI has been the exception — until now.

AISI is attempting to close that gap. Whether it's moving fast enough is a legitimate debate. But the work is real, the funding is real, and the output is beginning to appear.