Articles
Budget 2026: the final defence and AI decisions, and AIC’s response
Marketing and Outreach Team
7 Min Read
The final Budget documents are out. What changed after our initial briefing, which commitments predate today, and how AIC will look to turn relevant opportunities into tested capability.
Follow-up • 28 October 2026 • Sources checked at 14:28 GMT. The Treasury Budget document and the OBR’s October Economic and Fiscal Outlook are now public. This article replaces the provisional policy picture in our 13:01 briefing; the assessment of what AIC should do is our own.
Read the initial briefing, with its original 13:01 GMT source cutoff.
The economic constraint still matters
The OBR expects growth of around 1¼% in 2026 and 1% in 2027. Its central forecast puts inflation at a quarterly peak of 3¾% in early 2027, returning to 2% in 2028. Borrowing is forecast at £133 billion, or 4.2% of GDP, in 2026–27, falling to 1.8% in 2030–31. These are forecasts, sensitive to energy prices and interest rates, rather than guaranteed outcomes.
OBR: October 2026 Economic and Fiscal Outlook, executive summary and chapter 6.
Wider Budget decisions include £2.6 billion for household energy payments from January 2027, electricity VAT relief to July 2027, and a temporary bank surcharge increase from 3% to 6%. Headline income-tax, National Insurance and VAT rates remain unchanged.
For a software business, the practical implication is to cost delivery and support carefully. Customers facing pressure on their operating budgets need to see integration effort, infrastructure costs, maintenance responsibilities and the cost of leaving a system. A funded development trial can still become unaffordable to operate. Our bids and evaluations should expose that risk early.
June’s defence money must not be counted twice
Our first briefing identified the Defence Investment Plan’s £15 billion increase for 2026–27 to 2029–30, within roughly £298 billion of planned defence investment. June’s funding explainer identified £10.3 billion of funding and left £4.7 billion to be confirmed at this Budget. Those figures describe one package. The drone, targeting and transformation commitments discussed earlier were already within June’s plan.
The final Budget now says the plan is fully funded through spending reprioritisation. It targets 3% of GDP for defence in 2030 and 3.5% by 2035; the detailed trajectory follows at the 2027 Spending Review.
This answers the first question left open at 13:01: whether the previously announced plan would be funded. It does not justify adding the outstanding £4.7 billion to the £15 billion again. Nor should a spending target be treated as an open software competition.
Compare the original funding table: Defence Investment Plan funding explainer, 30 June 2026.
New finance and procurement decisions
New measures include a defence financing taskforce and up to £5 billion of National Wealth Fund/British Business Bank deployment over four years. A UK-supplier definition is due by January 2027; a new National Procurement Policy Statement follows in March.
Finance, investment and procurement serve different purposes. AIC will assess finance against the need it would meet, the terms attached and the revenue needed to sustain repayment or investment. We will assess procurement against a defined customer requirement. We should not plan on receiving a share of a headline allocation simply because we work in defence.
AI and cyber: what changed after 13:01
The 24 October Treasury announcement already contained £165 million for AI biosecurity, £120 million extra for the AI Security Institute, £60 million for Cyber Shield, including a £25 million challenge fund, and £70 million for the government’s NERV Centre. It also described a Manchester cyber and AI centre. Those were pre-Budget announcements, already covered in our first article; their appearance in the final document does not make them additional allocations.
The final procurement addition is £80 million, versus £60 million previously announced. AISI’s £120 million covers 2026–27 to 2029–30. The final document adds £40 million for an Agentic Security Unit and £24 million for public-sector hardening.
Treasury’s 24 October announcement provides the comparison baseline.
Two cautions: local AI partnerships are £30 million in chapter 5 but £15 million in chapter 7; the £25 million cyber challenge fund sits inside Cyber Shield’s £60 million. Delivery periods for several cyber allocations remain unspecified.
We will seek clarification of inconsistent figures before using them in a funding plan. The government’s NERV Centre and AIC’s NERVE are separate; the similar names imply no relationship or endorsement. A programme allocation also says nothing by itself about AIC’s eligibility, the date a competition opens or the contracting route.
AIC’s response: build the evidence before the bid
Our opportunity is to make difficult operational work more dependable. We will look for calls where our existing engineering and product work gives us a credible starting point, then agree a limited evaluation with the intended user. We should be able to explain the baseline, the proposed improvement, the test and the support model before choosing a funding route.
CORTEX and NERVE: trace an assessment back to its evidence
CORTEX supports intelligence analysis and lifecycle work; NERVE provides governed ingestion and search. We will look to evaluate the connection between source material, transformations and a reviewed assessment. A useful trial would ask an analyst to reproduce the reasoning, identify a changed source and check whether access restrictions still hold after reprocessing. Retrieval quality, correction effort and permission failures would be proposed measures, with results recorded rather than assumed.
That is a practical contribution to AI assurance: establish what information the system used and where a person made a decision. We would seek specialist partners where a challenge requires capabilities beyond ours. The biosecurity allocation, for example, does not make AIC a biological research organisation.
GTK: test imagery exploitation under realistic constraints
GTK is our geospatial intelligence and imagery exploitation toolkit. We will look for evaluations based on representative imagery, a defined analyst task and the intended deployment hardware. We would measure ingest and review behaviour, record unsupported inputs and test recovery after interruption. Where a customer needs local or disconnected operation, that requirement belongs in the acceptance test from the start.
PAL.NET: make integration and change easier to manage
PAL.NET is the Palantir Access Layer for C# and .NET. Our proposed work would focus on declared interface coverage, permissions, error handling and repeatable upgrade checks against the relevant upstream release. The test is whether a customer can connect the operations they need and maintain that connection. We will not infer universal API coverage or a particular security approval from a product name.
NAUTILUS: controlled development trials
NAUTILUS remains in development. We will look to scope bounded maritime trials around integration, observation, supervision and explicit control authority. Before any trial, we would agree operating limits, loss-of-link behaviour, manual takeover and responsibility for stopping the test. Development evidence must be kept separate from a claim of operational acceptance.
Cyber and agentic systems: enforce the boundary
For AI-enabled workflows, we will look to test constrained permissions, action logs, human approval and recovery. A permission check must still work when a model makes an unexpected request. An incident record should identify the input, proposed action, approval and actual effect. We would treat these as engineering requirements and test them with deliberately failing cases before proposing wider deployment.
What still needs an implementation answer
Before committing to an opportunity, we need the delivery body, call documents, opening dates, funding years, eligible activities, match-funding rules, intellectual-property terms and security requirements. We also need a customer who can provide lawful access to representative data and users, approve the evaluation and own the next stage. Announced spending cannot answer those questions.
For each suitable route, AIC will look to prepare a short evidence pack: the user problem, current product maturity, proposed work, acceptance criteria, cost, dependencies and route into supported service. A failed test should change the design or stop the proposal. That is how we intend to develop excellence through innovation: improvements a customer can inspect, reproduce and sustain.
Editorial note: this is AIC’s analysis of published policy, not an announcement of a government award, partnership or confirmed eligibility. Funding figures retain their stated scope; overlapping programmes are not added into an AIC market-size total.
Sources and earlier coverage
HM Treasury: Budget 2026 document, published 28 October 2026.
OBR: October 2026 Economic and Fiscal Outlook, published 28 October 2026.
AIC: initial Budget briefing, checked at 13:01 GMT on 28 October 2026.
Join our newsletter list
Sign up to get the most recent blog articles in your email every week.
Author
Marketing and Outreach Team
AIC’s Marketing and Outreach Team builds visibility and trust across Defence and security. We deliver strategic campaigns, thought leadership, and stakeholder engagement while balancing transparency with discretion. Our mission is to position AIC as a trusted, innovative partner to the UK MoD and beyond.
