Trojan horse fear as MoD uses American firm's AI to write review

Military chiefs' decision to use Artificial Intelligence as part a major strategy review poses major security risks, an expert said last night.

The Ministry of Defence's move to put an AI model - which was developed in the US - at the forefront of its probe into Britain's defence requirements could be exploited by rogue states and hackers, it was warned.

From the software used in fighter jets to delving into large datasets, AI technologies are expected to transform military operations in the coming years.

But it emerged yesterday that the MoD has deployed a custom-built AI programme to sift through submissions to its Strategic Defence Review, which will shape future decisions and spending priorities.

The Mail has campaigned for defence spending to be lifted to 2.5 per cent of Gross Domestic Product (GDP).

While the Labour Government has failed to set a timetable for increasing the budget to 2.5 per cent from the current spending ceiling of 2.2 per cent, it may do so based on the findings of the SDR.

But there are fears that using AI in the review could expose sensitive data held by the Armed Forces, according to senior MOD and NATO consultant.

Oxford professor Mariarosaria Taddeo, from Oxford University, who advises on digital ethics and regulation, said: 'Even if the AI is internal to an organisation, are we creating a Trojan horse?

'AI is really fragile, it can be attacked. How has the AI been developed, who's involved, what type of testing has been thought through?

'These are the key questions to consider whether AI-driven change of the defence sector is heading in the right direction.'

Read More

EXCLUSIVE
Military training cut back to fund pay rises: £400M shortfall forces MoD to scrap foreign exercises

The AI model being used as part of the SDR was created by US 'spy-tech' company Palantir Technologies.

The Silicon Valley firm has previously faced controversy over its NHS deal to upload patient info onto a central database, with critics raising concerns over confidentiality.

It is the first time the MOD has given AI such a prominent role in strategy and planning - and it comes as part of a broader push by the government to harness the technology.

Last night, defence sources told the Mail it was 'risky' to use an untested system on such an important project as the SDR.

A source said: 'The SDR will shape MOD decision making for the next decade, including spending.

'We need to be assured there is the appropriate human oversight of submissions to the review, it cannot be left to high-powered computers looking for key words.

'There is also the risk, as has been highlighted, that if Palantir is hacked or compromised, a third party could access some very sensitive documentation.'

The SDR will be led by former Nato Secretary General George Robertson and foreign policy expert Fiona Hill, a former aide to ex-US President Donald Trump.

With the assistance of AI, the findings of the review are expected to be published early next year. The conclusions will then form part of a MoD presentation to the Treasury to pitch for additional revenue.

Chancellor Rachel Reeves has ordered MOD to save millions of pounds in efficiencies – and to cover the costs of a pay-rise for troops.

The Mail recently revealed how the cost-cutting has including slashing the Army's training budget.

Defence officials also considered withdrawing £13 million in funding from the Royal Chelsea Hospital, home to the world-famous Chelsea Pensioners. But the plans were ditched following an intervention by the Mail.

Last night, the Ministry of Defence said: 'We have been transparent about our ambition to use AI for a wide range of defence applications.

'The team are utilising this technology to help review and analyse the high volume of submissions received by the SDR.

'We have consulted serving military, veterans, MPs, industry, academia and the wider public. Their submissions will be a key feature of the review.'