The list that confirmed it

The document that settled the question was not a policy paper. It was a list. On 20 July 2026 the government published Ministerial Appointments: July 2026, recording that the King had approved twenty-seven appointments as Andy Burnham took office as prime minister. Read the list looking for a technology secretary and you will not find one. Jonathan Reynolds appears as Secretary of State for Business, Innovation, Science and Trade, and no Secretary of State for Science, Innovation and Technology is named at all.

That absence is the news. The Department for Science, Innovation and Technology was created in 2023 under Rishi Sunak to give technology its own seat at the cabinet table, and it has now been merged into the business department, which takes the longer name. Speculation had run through the weekend. The appointment list is what turned it from a report into a fact you can act on.

What moved, and who owns it now

DSIT was not a small department, and it did not hold small things. It housed the AI Security Institute, the Sovereign AI Fund, UK Research and Innovation and the Government Digital Service, and it carried the AI Opportunities Action Plan. Those bodies were not abolished on Monday. Their reporting line was. All of them now sit under a secretary of state whose brief also covers trade, business and industry, with some functions understood to be moving to an enlarged culture and digital department.

For anyone with a live relationship to those institutions, this is a change of counterparty rather than a change of policy. A research grant from UKRI, a procurement running through the Government Digital Service, an evaluation commitment with the AI Security Institute, a consultation response you filed in June: none of those obligations disappeared, and all of them now belong to a different department with a different postal address, a different permanent secretary and a different set of competing priorities.

The objection was filed two days early

The sector saw it coming and said so in writing. On 18 July the Startup Coalition published a note titled plainly enough that no summary is needed, and it and techUK sent a joint letter to the incoming prime minister calling the abolition the wrong change at the wrong time and a terrible signal to a sector growing at roughly ten percent a year. Their argument was structural rather than sentimental: DSIT worked, they wrote, because it put researchers, AI practitioners, innovators and policymakers in one department and gave the sector a single front door.

Dom Hallas of the Startup Coalition put the risk in one line, warning that a mega department would mean British tech competing with British steel for attention. Matt Clifford, who has advised government on AI, called it a big mistake at a critical moment. The view was not unanimous. Tim Flagg of UKAI argued that a larger department could sharpen focus on AI rather than dilute it, while still warning about distraction and delay. Both sides of that argument, notably, expect a period of delay.

What a machinery change actually costs you

The part nobody writes down: a departmental merger has a cost that never appears in a press release, and it is measured in months of senior attention. Permanent secretaries, directors general and policy leads spend the first stretch after a reorganisation deciding who owns what, which budget lines transfer, and which teams sit where. During that stretch, decisions that require a signature from the top of the building slow down, and decisions that require a new department to form a view slow down further.

That is the exposure to price in. If your British roadmap contains a grant decision, a framework award, a regulatory clarification or a pilot with a public body, treat the timeline you were given before Monday as expired. It may be honoured. It should not be assumed. The safe planning posture is to carry the commitment at its current value while removing it from the critical path of anything that has to ship.

Where this leaves a 2027 plan

The institutions matter more than the letterhead, and they are intact. The AI Security Institute still evaluates frontier models, the Sovereign AI Fund still holds capital, UKRI still funds research. What changed is the level of political protection sitting above them. A programme with its own secretary of state has an advocate in every spending negotiation. A programme inside a department that also answers for steel, trade policy and industrial strategy has to win that argument internally before it is ever made in cabinet.

So the instruction for anyone building on British public infrastructure is narrow and specific. Keep the relationship, keep the funding line, keep the evaluation slot, and stop treating any of them as load-bearing for 2027. Build the plan so that it survives a reduced or delayed British commitment, and any continuity you get is upside rather than a rescue. Governments reorganise. Roadmaps that assumed they would not are the ones that break.