Statecraft At The Frontier

Cashing In Behind The Frontier

If you saw a technology developing that you believed could potentially change the trajectory of the world, for better or worse or somewhere in between, what would you do? Regulate it? Encourage growth and speed? Seek to use it for military purposes? These are decisions that the United States and China are able to ponder freely, thanks to their dominating artificial intelligence (AI) companies such as OpenAI and Anthropic in the West, and DeepSeek or z.ai in the East. Bar the ever-impressive Google DeepMind, Britain does not have these sorts of companies, so lacks the ability to steer the frontier. So what to do? Give up? Try and use ‘soft power’ to win commercial hearts and industrial minds? Talk about it forever on Britain’s only booming sector: waffling podcasts?

No. Instead, the British Government is developing a new form of technological statecraft. Unable to compete directly with the Americans and Chinese in building frontier AI, it has begun strategically establishing specialist institutions that give it insight into and occasional influence over the technology's development and application. The first part involves setting up government-backed organisations that can help Britain understand the technology, and potentially create ways of testing it that lead to safer and more secure developments. This is the purpose of the UK AI Security Institute (AISI), now copied globally, and widely viewed as the best in the world. It is also why the Government launched an AI Economics Institute earlier this year to understand the potential impact of this technology on the economy and labour markets. But while being able to look under the hood of these frontier technologies is essential, it alone is not going to be enough to insulate Britain from negative effects, nor commercialise some of the benefits.

As a pioneer of microbiology and vaccination Louis Pasteur observed over a century ago, “without laboratories, men of science are soldiers without arms.” This then leads to the second act: the launch of the world’s first government-backed AI Biosecurity Lab (AIBL). The Government is specific: AIBL “is designed to establish the UK as the world-leader in AI-enabled biosecurity.” This moves Britain’s strategic gamble from analysis and high level testing to applications at the intersection between AI and biology with two ideas in mind: help protect Britain from future pandemics (including potentially AI-created or enhanced ones), and help develop a biology industry at the frontier of this technology, tapping into deep existing strength across Government and the private sector.

Can I shock you? There are risks and opportunities

Biology is not unlike government. It can take many forms, behave in ways we simply cannot understand without further deep study, and wreak total havoc on society if left to its own devices. What is less complex is the regularity of pandemics. These tend to occur every few decades, and - as we saw with Covid-19 - can cause appalling loss of life and retardation of economies and societies around the world if not dealt with. The wonks at the World Bank actually model the probabilities of major pandemics: their assessment is that “the risk…rises over 6 per cent annually, driven by deforestation, agricultural intensification, urbanization, climate change, and growing human-animal contact.” Cheerfully, they conclude that “another pandemic on the scale of COVID-19 has close to a 70 percent chance of happening within the next 25 years.” This is before you add in any potential impact from AI, which the Government summarises as: “creating unprecedented opportunities for scientific discovery, but…also changing the biological risk landscape [because] AI can accelerate research, improve our understanding of diseases, and support the discovery and development of new medical countermeasures. However, increasingly capable AI systems also create new biological risks and could lower barriers to harmful activity.” And as the head of MI5 recently said in his annual speech: “the misuse case to which I’d attach greatest weight is AI’s potential to change the stakes on existing or even new bioweapons.”

Consider the following three points. Half a decade ago ago, the British Government stated publicly that “It is likely that a terrorist group will launch a successful [chemical, biological, radiological or nuclear] attack by 2030.” In the intervening years, frontier AI companies such as Anthropic, OpenAI and Google DeepMind have warned that their technologies may make creating bioweapons easier, and have sought to put guardrails to limit this possibility. And earlier this year, an analyst’s interviews with over two dozen former Boko Haram members revealed an appetite for using AI systems to improve fighting abilities, embedded ‘AI units’ in their operations, and an interest in potentially applying the technology to the creation of bioweapons should the opportunity arise. One terrorist told the analyst that while “unless you can invent or buy them, you waste your time discussing this”, they would not answer any questions as to if they had tried to use the technology to invent new bioweapons.

Now consider three other points. Half a decade ago, AI enabled Google DeepMind to create a breakthrough in the shape of AlphaFold, helping predict the structure of nearly all catalogued proteins known to science. In 2023, scientists used machine learning to find a molecule, since called abaucin, that kills Acinetobacter baumannii, a dangerous pathogen often found in hospitals that was viewed as resistant to antibiotics. And late last year, researchers demonstrated that AI could design entirely new antibodies intended to attach to specific parts of disease-related proteins.

The AIBL represents a strategic effort by the British state to pre-empt and pre-act on this complex landscape, spotting both the concerns as they arise and, hopefully, creating some benefits as they explore and research this frontier. Its purpose is to both prepare for a crisis scenario by having enhanced early detection of emerging biological threats and to accelerate the development of vaccines, treatments and other medical countermeasures. In doing so, it will be working closely with the deep expertise already found in Britain’s biosecurity landscape and the private sector.

Notably, AIBL has been set up as an independent nonprofit, allowing it to recruit technical talent from industry and academia and operate with considerable autonomy, whilst still working with the pre-existing government agencies. This is helpful because as it currently stands, the market doesn’t tend to produce adequate pandemic preparedness on its own, whereas the state has both the incentive and the authority to invest in capabilities whose benefits are dispersed across society. The partnerships with the private sector is another critical element: OpenAI is committing £100m in compute and model credits through its Rosalind Biodefense programme, Anthropic up to £100m in model credits, access and technical support, and Google DeepMind and British firm Basecamp Research contributing expertise too.

A New British Export

It should be clear: Britain is not going to outspend Silicon Valley or outbuild China's AI champions. The AIBL is not an effort to do so. Instead, it is an example of taking a strategic decision to assemble institutions with access to the world's most capable models, the technical talent to deeply analyse them and spot emerging patterns and behaviours, and a driving mandate to pursue problems which neither venture capitalists nor government departments are particularly well equipped to solve. Should it succeed, it provides two further notches for the British state’s bow: a potentially useful tool in the US bilateral relationship, and a tangible effort that a Government can point to when its population asks what it is doing to deal with these potentially deadly threats.

We are rarely blessed with the ability to understand how critical decisions made today will prove in a decade's time. Britain's AISI, founded three years ago, has already helped inspire similar institutions around the world, and the AIBL may prove equally influential. This is fantastic, but it is not decisive. The proof will be in the pudding: Britain will need to turn its early lead into technical expertise, scientific breakthroughs and relationships into tangible outcomes when the time comes.