TL;DR
The Ministry of Defence is building a targeting system described in its own slides as “autonomous by default, human by exception”, according to documents the Bureau of Investigative Journalism (TBIJ) obtained under freedom of information and reported on 13 November. The system, part of the billion-pound Digital Targeting Web programme, could be deployed as soon as next year. TBIJ also found that the MoD’s AI Ethics Advisory Panel last met in December 2025.
What the slides say
The material comes from a presentation given this year to prospective industry partners. TBIJ reports that a first version should be able to pick out hundreds of targets weekly from next year, while the 2030 ambition appears to be tracking thousands of “dynamic targets” at once.
Two lines stand out. One says the MoD wants new AI weapons technology “into the field as soon as possible”. Another, also cited by the defence think tank RUSI earlier this year, reads: “Assurance, legalities, ethics and security must accelerate tempo, not constrain it.” RUSI said that approach “fundamentally misunderstands the purpose of safeguards”.
The MoD gave no reply to repeated TBIJ requests for comment. The Foreign Office said it “will always comply with international humanitarian law” and is “actively engaging in international discussions”, but would not state its position ahead of UN talks in Geneva on autonomous weapons, which TBIJ, writing on 13 November, said would open the following week.
Oversight gaps
The ethics panel, set up in 2021, was presented by the MoD as central to responsible military AI. Its intended successor, an AI Expert Advisory Group, has yet to hold a first meeting, the MoD told TBIJ.
Richard Moyes of Article 36, a non-profit working on weapons law, called the “accelerate tempo” principle “genuinely shocking”, adding: “It is hard to read this as anything other than an explicit downgrading of legal and ethical considerations, which are supposed to provide guardrails not accelerators.”
The Red Cross (ICRC) urged states, including the UK, to move from agreed elements to negotiating a protocol “before it is too late”. A diplomat involved in the talks told TBIJ that removing a human from the decision to kill makes international humanitarian law “very complicated”.
Looking Forward
The disclosures arrive just before Geneva, where the US, Russia, Israel and a handful of other states have argued against binding rules. In our view, the gap between the MoD’s public language on responsible AI and the tempo-first wording in its supplier pitch will be hard to reconcile without either a published oversight framework or a convened advisory body. We would expect suppliers bidding into the programme to face parliamentary questions on both.