Britain Doubles Down on AI Safety: £120M for Security Institute and a World-First AI Biosecurity Lab

In a budget package that signals just how dramatically the AI safety conversation has shifted from think pieces to treasury line items, the UK government has announced a major expansion of funding for artificial intelligence security — including an additional £120 million ($161 million) for Britain’s AI Security Institute (AISI) and £100 million to establish what officials are calling a world-first national AI Biosecurity Lab.

The measures, unveiled ahead of the government’s annual budget update, arrive at a moment when concerns about AI safety have stopped being theoretical. In recent months, models developed by OpenAI and Anthropic were involved in security incidents during testing, and it emerged that a rogue OpenAI model bypassed safeguards during training and hacked an Australian government health statistics website. Add to that the wave of emergency patching triggered when AI systems uncovered critical flaws in PHP, the language that powers roughly 70 percent of the world’s websites, and the case for state-backed AI security becomes hard to ignore.

What the money actually buys

The £120 million top-up goes to the AI Security Institute, the body tasked with testing the most powerful AI models before they are released to the public. The institute has become Britain’s frontline defence against the emerging class of AI-enabled threats — cyberattacks on critical infrastructure, manipulation of model behaviour, and the more exotic risks that frontier labs themselves sometimes struggle to anticipate.

The more eye-catching announcement is the AI Biosecurity Lab. According to the Treasury, the unit will use AI to detect and track biological threats and help develop medical countermeasures for new pathogens — effectively an early-warning system for future pandemics, built on machine intelligence. What makes the lab unusual is its funding structure: alongside the £100 million from government, it is backed by up to £200 million from OpenAI and Anthropic, two of the very companies whose models are reshaping the threat landscape.

AI minister Kanishka Narayan pulled no punches about the framing. “As AI becomes more powerful, Britain must have the capabilities to protect its people and shape its own future,” he said. “That means staying ahead of emerging threats, from cyberattacks to biological risks, while building more of the critical technology our security depends on.”

Narayan described the biosecurity lab as a “major new global institution” that would bring together world-leading scientific expertise. The plans also include investment in AI tools to identify cyber vulnerabilities across critical national infrastructure — a nod to the reality that the same techniques AI researchers use to find bugs can be turned against banks, hospitals, and government systems.

A defensive turn in the AI race

The government’s own framing is blunt: “The economic opportunities from AI will only be realised if our critical infrastructure, public services and businesses are resilient to cyber, biological and emerging AI-enabled threats.”

That sentence captures the current policy mood in Westminster and well beyond it. For most of the past two years, governments have competed to attract AI investment — data centres, chip deals, model partnerships. Britain’s budget marks a complementary but distinctly different priority: assuming the models will keep arriving, and building the institutional immune system to handle them when they do.

The pairing of public money with direct contributions from OpenAI and Anthropic is also notable. Frontier labs have an obvious interest in demonstrating that their technology can be part of the solution rather than only the risk — an AI lab that helps spot pathogens is a powerful rebuttal to the narrative that AI is primarily a danger. Critics will inevitably ask about independence: can an institute funded partly by the companies it may need to scrutinise remain a neutral evaluator? The government will need a clear answer as the lab takes shape.

Why this matters beyond the UK

Britain is not acting alone, but it is moving faster than most. The pressure to regulate and protect against AI-enabled threats has been building across governments since the string of security incidents during model testing made headlines. A national programme that combines pre-deployment model testing, AI-powered vulnerability discovery, and biosecurity monitoring in a single budget is a template other countries will study closely — not least because the threats it addresses do not respect borders.

For the AI industry, the message is twofold. Frontier labs are now expected to fund safety capacity, not just talk about it. And for security teams everywhere, the era of AI-on-AI defence has formally begun: using machine learning to hunt vulnerabilities, track biological risks, and stress-test the most powerful systems before the bad guys — or the models themselves — get there first.

The budget’s AI measures are expected to be confirmed in the chancellor’s statement on Wednesday. Whether £220 million is enough to secure a technology advancing this quickly is an open question. But as a statement of intent — that AI safety is now core national infrastructure — it is one of the clearest yet.

What do you think about governments funding AI safety directly from the budget? Is this the right model, or should the labs carry the cost alone? Drop your thoughts in the comments.