Britain must not rely on foreign AI, a former military intelligence chief has warned.
Gen Sir Jim Hockenhull, who served as head of Cyber and Specialist Operations Command (CSOC) until June, called on the Government to exercise extreme caution when relying on AI contracts obtained abroad at a time when Russia is becoming more aggressive in its hybrid attacks on Britain.
He said: “The world has rarely felt less safe.”
In his first comments since leaving the Armed Services after a 40-year career, Gen Hockenhull wrote in The Telegraph that hostile states were launching attacks through cyberspace, disinformation, sponsoring terrorism and conducting espionage and sabotage “on a scale not seen before”.
He warned that while AI will “redefine warfare” and provide many benefits for Britain, including affording military analysts “greater speed and accuracy”, it can also “leave the UK more exposed to cyber-attack”.
“A technology that helps the defender also helps the attacker,” he wrote, as he stressed the need to understand any AI models that are used in military planning and intelligence analysis.
The Telegraph revealed in September that over the past five years more than £30bn of UK defence contracts have been awarded to foreign suppliers, despite being counted as domestic spending.
Thousands of contracts that ministers claimed had been awarded in the UK had, in fact, been given to companies headquartered offshore, with data obtained under freedom of information laws that show that since 2021, ministers have awarded £31.7bn in defence contracts to foreign-controlled firms with UK offices, compared with £3.5bn spent with foreign suppliers.
The four-star general, who has been made president of the British technology startup Valarian, warned that Britain’s dependence on foreign AI was deepening.
“Five companies control 70 per cent of global AI compute, none of them British,” Gen Hockenhull said.
“We must not be a high-value AI customer. A state that runs its tax system, its intelligence analysis and military planning on models owned, hosted and updated abroad is not master of its own house.
“It is a tenant, and tenants can be evicted, repriced or spied upon. This is a critical vulnerability, especially so in time of crisis.”
As head of CSOC, previously called Strategic Command, Gen Hockenhull held one of the most senior jobs in the Armed Forces, overseeing UK special forces, Defence Intelligence, National Cyber Force and Permanent Joint Headquarters.
In this role, he saw up close how hostile states were “attacking infrastructure through cyberspace, running information operations, sponsoring terrorism and conducting espionage and sabotage on a scale not seen before”.
“Modern conflict is no longer confined to the battlefield; it is not constrained by lines on maps,” he said.
“It runs through cables, across the electromagnetic spectrum, in data centres and supply chains. And it is under way long before anyone declares war.”
Gen Hockenhull said four things were required to regain control of the situation, which included the Government inspecting and testing any systems it uses, ensuring sensitive data sit on infrastructure it trusts, not being held “hostage” by relying on single suppliers, or foreign governments, and had to have the skills to be self-sufficient.
He said that Britain should not be timid in its approach to working with AI, but it had to “own the keys” while doing so.
Britain must own the keys to AI
by Gen Sir Jim Hockenhull
The world has rarely felt less safe. Those who run Britain’s Armed Forces and Civil Service, usually a cautious lot in public, now speak in unusually plain terms about the threats we face. Hostile states are attacking infrastructure through cyberspace, running information operations, sponsoring terrorism and conducting espionage and sabotage on a scale not seen before. Modern conflict is no longer confined to the battlefield; it is not constrained by lines on maps. It runs through cables, across the electromagnetic spectrum, in data centres and supply chains. And it is under way long before anyone declares war.
Into this world a technology of startling promise has arrived. AI will transform our lives. It will redefine warfare and may prove the deepest shift in how states are governed. The evidence is already visible. HMRC has used machine learning to protect and recover £10bn in a single year, and AI is helping to halve decision times on routine householder planning applications.
The same logic applies to defence. AI gives military analysts greater speed and accuracy, sifting surveillance data that no human team could read. It helps make faster and better decisions and is vital in augmenting other new technologies. In a nation with tight budgets, a shrinking pool of talent and a long list of threats, a technology that multiplies the output of each person must be exploited.
But a technology that helps the defender also helps the attacker. Officials have been candid that an AI-enabled state could leave the UK more exposed to cyber-attack. The more Whitehall rebuilds its workflows around AI, the more those systems become a prize. Defence chiefs add that adversaries will use AI to degrade British capabilities, and that the military must understand how AI is working when it enters decision-making chains. A model that cannot be explained cannot be trusted as part of life-or-death judgment, and one that cannot be audited cannot be trusted with a welfare payment either.
Here sovereignty stops being a slogan. Britain’s dependence is deep and getting deeper. Five companies control 70 per cent of global AI compute, none of them British. We must not be a high-value AI customer. A state that runs its tax system, its intelligence analysis and military planning on models owned, hosted and updated abroad is not master of its own house. It is a tenant, and tenants can be evicted, repriced or spied upon. This is a critical vulnerability, especially so in a time of crisis.
Sovereignty need not mean autarky. The Government is right to stress that it is not a call for isolationism or for going it alone. Britain will never build every chip or train every frontier model, and it would be foolish to try. The aim is narrower and more achievable: control over what the country cannot afford to lose. That means four things. The Government must be able to inspect and test the systems it uses. Its sensitive data must sit on infrastructure it can trust; this is equally true for any company or organisation that wants to retain control. It needs credible alternatives, so no single supplier, or foreign government, can hold it hostage. And it must have the skills to do all this itself.
The last necessity is the scarcest. A new AI-focused intake for the Civil Service’s graduate scheme is welcome. But Whitehall alone cannot buy its way to sovereignty, and it cannot hire its way there on Civil Service pay scales alone. It will need partnerships with British firms, universities and like-minded allies, and the nerve to back domestic champions before they are obviously safe bets.
Nothing here argues for caution about AI. The costs of timidity are as real as the risks of haste: slower services, weaker defences and a nation that lags behind our competitors and our adversaries. But speed without control is merely dependence at a higher tempo. Britain should adopt AI boldly. It must insist, just as boldly, on owning the keys.
Gen Sir Jim Hockenhull is a former commander of Cyber and Specialist Operations Command and is now president of Valarian Technologies