AI's Overhaul of Government

AI and The Imminent Obsolescence of Analog Government

“Without Artificial Intelligence, states will fail. The increasing complexity of modern life has transformed every aspect of our existence. The simple example of the pencil, famously used by Milton Friedman to illustrate interconnected systems, now applies an exponential way to every part of society.

It is no longer possible to operate a healthcare system - or even something as straightforward as a road repair network - without the support of AI. The choice is between adapting for complexity or for chaos.”

The most dangerous form of optimism is procedural. That unflinching, institutional belief that if we just refine the process, if we rearrange the committee chairs with sufficient finesse, the future will wait patiently for us to be ready. Ed de Minckwitz’s brilliant new report: Government in the Age of Superintelligence detonates this illusion with stark clarity. The report isn’t alarmist, but it should be alarming, a sharp, cerebral briefing for politicians still pretending they have more time than they do.

The report indicates that we are not witnessing a technological shift. We are facing a civilisational inversion.

“The Dawn of a New Era,” the report's prelude, doesn’t offer forecasts. It offers obituaries in advance: for linear thinking, for analog governance, for the fiction that institutions built for ration books and rotary phones can adapt, in time, to a force that rewrites the meaning of time itself. Superintelligence is not a tool. It is a recursive force multiplier whose substrate is cognition. It doesn’t just accelerate history, it compacts it. As Dario Amodei calls it, a “compressed century” looms, where a hundred years of social, economic, and epistemological transformation happens between two electoral cycles.

Slow Demise

Traditional governance, de Minckwitz argues, is already outpaced by narrow AI. As he observes, the civil service functions like a “primitive large language model, with all the hallucinations and bias that they exhibit,” producing policy based on historical precedent in a world where the future is unrecognizable.

Ministers are trapped in a cycle of crisis management, with de Minckwitz suggesting none spend more than five percent of their time on AI, and not one has it as their sole portfolio. Not one. In this vacuum, governments continue drafting long-term economic strategies and NHS roadmaps without accounting for technologies that will, in a handful of years, redefine what jobs exist, how diagnosis occurs, and even how people age.

The anthropocentric delusion, the assumption that human intelligence remains central, crumbles under the weight of systems that have already outstripped our capabilities in pattern recognition, legal synthesis, and scientific modeling. What does it mean for governance when the average policymaker is no longer the most intelligent agent in the room? When even the most elite knowledge work becomes replicable, then surpassable, by agentic systems? De Minckwitz doesn't speculate; he recounts. The Department for Work and Pensions, at one point, banned the use of ChatGPT among staff. In effect, the government disabled access to a calculator during the dawn of calculus.

The report proposals, a Prime Minister’s Superintelligence Council, dedicated AI ministers, and an independent Office for Superintelligence, are not rhetorical provocations. They are deliberately modeled on wartime and climate structures: frameworks that assume radical uncertainty and demand proactive coordination. The idea is not just to prepare for AI, but to re-architect governance so that it becomes capable of strategic anticipation.

New Roles

Sector by sector, de Minckwitz sketches what the future will demand:

In healthcare, he foresees a shift from reactive treatment to predictive diagnosis, with the possibility of curing once-intractable diseases and perhaps extending human longevity by fundamentally reshaping human biology itself. What happens to the NHS's ten-year strategy when diagnosis, prescription, and recovery can be accelerated a hundredfold by systems that simulate molecular interaction faster than any lab?

In labour, he flips the conventional wisdom. Roles requiring manual dexterity, empathy, and real-world responsiveness may become the new premium, while white-collar professions like law and finance, long considered ‘high-skilled’, face obsolescence. What then becomes of a society whose credentialing systems reward the soon-to-be replaceable?

Upping the Tempo

In defence, the threat isn't just automated weaponry but strategic instability. De Minckwitz warns of a world where an AGI arms race incentivizes preemptive strikes on research labs, not missile silos. This isn't cold war logic; it's a hotter, faster logic in which escalation is algorithmic.

And yet, the most audacious move in the report is its insistence that we still have agency. Governments, de Minckwitz insists, must begin using AI to prepare for superintelligence. Not in a vague, efficiency-driven sense, but in structured ways: for scenario planning, stress-testing of policies, legislative foresight, and public communication. We are, in his words, failing to use the very thing that could save us from what comes next.

But does the report go far enough? One might ask whether its solutions, bold as they are, are implementable at the scale and speed required. Can Whitehall, where it takes a decade to build a bridge or revise a pension scheme, rewire itself fast enough to meet the speed of recursive improvement? Even the recommendation for superintelligence impact assessments presumes a policy tempo that current departments seem structurally incapable of achieving. De Minckwitz admits this. Root-and-branch reform, he says, is rarely achievable. But he proceeds anyway, perhaps aware that the alternative is not stasis but irrelevance.

And there remains the alignment problem. De Minckwitz assumes, for the sake of argument, that we will solve it. But is this not a subtle return to the very procedural optimism he critiques? We may hope to embed human values in superintelligent systems. But we have yet to reach consensus on what those values are. Worse, we lack the institutional literacy to translate them into code.

What the report demands, ultimately, is not another strategy document but a civic and cognitive overhaul.

“whether we arrive at the Age of Superintelligence in five years or fifty, the magnitude of its potential impact demands immediate preparation. Governments cannot afford to wait until the technology is fully realised before beginning to adapt and adopt.”

A state that governs in the age of superintelligence must be able to think probabilistically, to legislate recursively, to build systems that learn. It must swap hierarchy for modularity, retrospection for simulation, and political rhetoric for executable foresight. The premise of this Policy Exchange report is that superintelligence will demand a fundamental reimagining of governance structures.

De Minckwitz has written something rare: a document that straddles the line between statecraft and philosophical provocation. It does not pretend to have all the answers. But it is unflinching in its claim that the old questions no longer suffice.

The superintelligence is coming. The only question left is whether the fax machine is still in use and who gets to reply. In other words, will any politician be listening?

Stay curious

Colin

It is also worth watching this recent 18 minute video with Sir Demis Hassabis who covers many of these points.