The UK Gives AI a Cabinet Seat While the US Army Rations It
Two stories about governments and AI, published hours apart, point in opposite directions. In the UK, Prime Minister Andy Burnham has abolished the Department for Science, Innovation, and Technology outright, and elevated AI Minister Kanishka Narayan — who also oversees the AI Safety Institute — to a full seat at cabinet. In the US, an Army CIO memo reinstates limits on how many AI tokens Army personnel can use, even as internal documents show the Department of Defense’s own Ask Sage tool has access to 100 million tokens through an annual “enterprise pack.”
Restructuring Around AI vs. Rationing It
Abolishing an entire department to elevate a single AI-focused minister to cabinet is a structural bet: it says AI policy is now important enough that it shouldn’t be bottled inside a broader science-and-technology portfolio, and that whoever runs it needs a direct line into the highest level of government decision-making. Narayan’s oversight of the AI Safety Institute means the person now sitting in cabinet is also the one responsible for the UK’s frontier-model safety evaluation work — pairing policy authority with the technical safety function in a single seat.
The Army’s move is the mirror image. Rather than restructure around AI, the Army is pulling back on how much of it its own people can use day to day, reinstating token limits that had presumably been loosened at some point. The detail that the DOD’s Ask Sage platform holds access to 100 million tokens annually, while individual soldiers face renewed caps, is the real story inside the story: this isn’t the US military rejecting AI, it’s an institution trying to control consumption and cost at the individual-user level while keeping a much larger allocation available at the enterprise layer. That’s a rationing problem, not a rejection problem — the kind of thing that shows up when token spend scales faster than a budget line was built to absorb.
Same Technology, Two Institutional Postures
Neither government is against AI. The UK is reorganizing its cabinet to put a person in charge of it; the US Army is metering how its people use it. Both are responses to the same underlying fact — that AI has become significant enough, and expensive enough, that the old organizational defaults no longer fit — but they land on opposite sides of the same problem. One government concluded the answer was to give AI more institutional authority. One branch of the other concluded the answer was to give individual users less operational latitude. Watching which posture other governments and other branches of the US military converge toward over the next year is a reasonable proxy for how the AI-and-government relationship is going to settle.