Is Britain Ungovernable?

THE UNGOVERNABLE STATE?

Britain, the AI Century, and the Compounding Crisis of Capacity

A strategic analysis  ·  17th May 2026

Preamble: A Diagnosis and its Implications

The claim that Britain has become ungovernable is now made openly — by historians, by foreign correspondents, by its own politicians, occasionally by its own Prime Ministers. It deserves to be taken seriously, not as a counsel of despair but as a diagnostic tool. If the diagnosis is substantially correct, it has consequences that go far beyond electoral politics. It touches the country's capacity to defend itself in an era of AI-enabled warfare, to resist long-running foreign penetration of its institutions, to manage the social fractures that entitlement culture and unresolved immigration policy have deepened, and to participate meaningfully in the multilateral governance structures that the AI century requires.

This paper argues that the 'ungovernable' thesis, while overstated as an absolute claim, identifies something real and dangerous: a compounding failure of institutional capacity, political imagination, and strategic literacy that has been accumulating for decades and is now colliding with the most demanding technological and geopolitical environment Britain has faced since 1940. The collision between an ancien régime political system and the realities of the AI century is not a political inconvenience. It is a structural vulnerability.

Britain is not ungovernable. But it is being governed by people, through institutions, and with ideologies designed for a world that no longer exists — in an era that will punish that mismatch without mercy.

Part I: The Political System and its Exhaustion

Six prime ministers, one crisis

Since the 2016 Brexit referendum, Britain has cycled through six Prime Ministers: May, Johnson, Truss, Sunak, and now Starmer — with a seventh apparently in preparation as Labour's internal collapse accelerates. To put this in perspective: it took Britain more than forty years to go through that many leaders in the period before 2016. The revolving door is not merely a symptom of individual failure. It is evidence that the system has lost its capacity to generate stable political authority.

The immediate picture is stark. As of early 2026, nearly three-quarters of voters believe things are getting worse, and only 8% think they are improving. Economic optimism is at its lowest recorded level since 1978. Starmer's net favourability rating of minus 57 matches only Liz Truss — a Prime Minister who lasted 49 days. Reform UK is projected, on current polling, to win a parliamentary majority with 381 seats. Labour, which won the largest parliamentary majority since 1997 less than two years ago, is projected to fall to 85 seats. These are not the numbers of a political system under strain. They are the numbers of a political system in decomposition.

A first-past-the-post system in a five-party country

The structural cause is not difficult to identify. Britain's first-past-the-post electoral system was designed for, and functions adequately in, a two-party political environment. It is structurally unsuited to the five-way fight — Labour, Conservatives, Liberal Democrats, Greens, and Reform — that now characterises English politics, with additional complexity in Scotland and Wales. The system's response to multi-party fragmentation is not stable government but perverse outcomes: large majorities for parties with minority support, total collapses for parties with substantial vote shares, and parliaments that do not reflect the preferences of the electorate.

The consequence is not just representational unfairness. It is a crisis of legitimacy. When a government elected with a large majority on a minority of the vote then governs badly, loses the confidence of three-quarters of the population within eighteen months, and is replaced by a party that the same electorate did not choose, the democratic contract is functionally broken. The institutions survive, but the animating belief that they represent the people does not.

Ideology without reality

What makes this political crisis particularly acute in the context of the AI century is the ideological character of the parties now contesting power. Labour, in its current form, combines social-democratic economic instincts with a lawyer's proceduralism and a managerialist faith in process — none of which constitutes a coherent response to the strategic disruptions described in this paper. The Conservatives, in their various recent iterations, have alternated between libertarian market ideology, culture-war populism, and a nostalgia for a global Britain whose material basis was dissolved by deindustrialisation and Brexit. Reform offers grievance politics and nationalist instinct with no serious policy architecture behind it.

None of the major parties has produced a coherent account of what the AI transformation means for the British state, its economy, its security, or its social contract. None has seriously engaged with the question of hyperscaler dependency. None has addressed the strategic implications of a decade of accumulated cyber penetration by hostile states. The debate about Artificial Intelligence in British politics is conducted almost entirely at the level of jobs anxiety and regulatory optics. The deeper questions — about sovereignty, resilience, the distribution of AI capability, and the fitness of democratic institutions to govern at machine speed — are absent from the political conversation.

The parties argue about who should manage the decline. None has a serious account of what would reverse it. In an AI century that demands strategic literacy as a basic governing competence, that absence is not a policy gap. It is a disqualification.

Part II: The Civil Service and the Capability Deficit

Whitehall in the AI era

The British civil service was, for much of the twentieth century, among the most admired bureaucratic systems in the world: generalist, politically neutral, intellectually serious, and resistant to corruption. That reputation has not survived the last two decades intact. The 'Yes Minister' caricature of a service that manages ministers rather than serves them has become less satirical and more diagnostic. But the more serious problem for the AI century is not the political culture of Whitehall — it is the technical capability gap.

The evidence is unambiguous. The former head of the government's AI incubator has said plainly that not enough technical people work in the civil service and that civil servants do not understand technologists or technology. The 2026 State of the State report found that nearly 90% of public sector workers admit their organisation cannot fully leverage AI today. The NCSC handles around four nationally significant cyber incidents per week, a rate that has more than doubled year on year — yet the civil servants responsible for commissioning and overseeing the government's digital infrastructure routinely lack the technical literacy to understand what they are buying, from whom, or what dependencies they are creating.

The government's own AI pilots — the Humphrey suite, including Redbox, Parlex, Caddy, and Medguard — have been closed or discontinued, with the stated reason that commercial alternatives now outperform them. This is presented as a pragmatic decision. It is also, in strategic terms, a surrender: the government has acknowledged that it cannot build the tools it needs and must instead buy them from the same private sector whose growing dominance over public infrastructure this paper has identified as a structural risk.

The shadow state problem

Civil servants using personal ChatGPT accounts to perform their jobs — a trend noted without apparent alarm in official reports — are sending government data through systems with no oversight, no audit trail, and no data protection controls. This is not an edge case. It is a description of how significant portions of the British state are currently functioning: dependent on foreign private AI systems, operating outside their own data governance frameworks, and creating vulnerability vectors that a capable adversary can exploit.

The gap between what the civil service knows about AI and what it would need to know to govern it responsibly is not a training problem that can be solved with a playbook and some online courses, however well-intentioned. It is a structural mismatch between the kind of institution Whitehall is — generalist, cautious, slow to hire specialist skills, oriented toward process compliance — and the kind of institution that governing an AI-era state requires. That mismatch will not be resolved by the current reform programme, which plans to cut civil service headcount by 15% while using AI to 'boost efficiency' — a proposition that assumes the problem is too many civil servants rather than too few with the right capabilities.

The British state is trying to govern an AI century with a civil service built for the industrial one — and is responding to the gap by reducing the size of the civil service rather than transforming its capabilities.

Part III: The Accumulated Damage of Hostile State Penetration

A decade of undeclared war

The head of the NCSC warned in April 2026 that the UK is living through 'the most seismic geopolitical shift in modern history' and that British businesses must prepare to be targeted at scale in the event of conflict. This warning would carry more weight if it were not delivered against a backdrop of more than a decade of documented, systematic penetration of British institutions by Russian, Chinese, and Iranian state actors — penetration that has continued largely uninterrupted, and whose cumulative effects have not been publicly accounted for.

GRU Unit 29155, exposed in a joint advisory with nine allied nations, has conducted sustained cyber operations against government and critical infrastructure organisations globally, including in the UK. GRU Unit 26165 — APT28, Fancy Bear — deployed malware specifically designed to enable persistent, long-term access to Microsoft cloud accounts, stealing credentials and authentication tokens while concealing its activity from the victim's sent folder. The NCSC has handled more than 200 nationally significant incidents in a single year. These are not isolated intrusions. They are elements of a sustained campaign of strategic intelligence collection, infrastructure mapping, and institutional degradation.

What sustained penetration actually means

The political and media conversation about hostile state cyber activity focuses on incidents: the hack, the breach, the exposed dataset. This framing systematically understates the strategic significance of what is happening. The purpose of sustained, long-term penetration of government systems is not primarily to steal specific documents. It is to build a persistent, evolving understanding of how British institutions work, where their vulnerabilities lie, who the key decision-makers are and what their relationships and pressures are, and how to manipulate or disrupt those systems at a moment of the adversary's choosing.

A hostile state that has had persistent access to Foreign Office communications, to Cabinet Office systems, to defence procurement databases, to the systems of critical national infrastructure operators, for years or decades, possesses something more valuable than any individual secret: a living map of the British state's decision-making anatomy. This is the information environment within which Project Mythos would need to operate. It is the information environment within which Britain's political class is making decisions about AI, about defence, about economic policy. The question of whether that environment has been substantially compromised is not one that current political discourse is asking.

The Mandelson affair — in which the British Ambassador to the United States was denied security clearance by UK Security Vetting in January 2025, a decision overridden by officials under political pressure, and in which the Foreign Secretary subsequently dismissed a senior Foreign Office official who raised concerns — is illustrative not because it proves espionage but because it demonstrates the institutional fragility around security culture at the highest levels of government. When security vetting recommendations can be overridden for politically convenient appointments, and when the officials who raise concerns are removed, the signal sent throughout the system about the relative priority of security versus political convenience is unambiguous.

The question is not whether British institutions have been penetrated by hostile states. The question is whether the accumulated effect of that penetration is now shaping British political decisions in ways that neither the public nor most politicians are aware of.

Espionage as long-term structural damage

There is a version of the 'ungovernable' thesis that has nothing to do with electoral politics or ideological failure and everything to do with the slow, invisible degradation of institutional competence and trust that sustained intelligence penetration produces. Organisations that have been penetrated develop cultures of caution, compartmentalisation, and internal distrust that reduce their effectiveness. Decision-making becomes slower and more risk-averse. Key relationships are managed with one eye on potential exposure. The informal information networks through which effective governance actually operates — the conversations that happen before the formal meetings, the honest assessments that never appear in official documents — become vulnerable to exploitation.

Britain has been the target of sustained hostile state intelligence activity for decades. The Russian operation to cultivate and recruit assets in British political, academic, and media circles — documented in the Russia Report that the Intelligence and Security Committee had to fight to publish in 2020 — did not begin in 2016 and has not ended. The Chinese infiltration of British universities, research institutions, and technology companies has been the subject of repeated warnings by GCHQ and MI5 that have produced inadequate policy responses. The cumulative effect of this activity on the quality of British institutional decision-making is not something that can be measured or publicly attributed. It may nonetheless be significant.

Part IV: Immigration, Entitlement, and the Social Contract

The Hermer/Starmer dynamic

The paradox of the Starmer government's approach to immigration is that it has managed to be simultaneously more restrictive in rhetoric and more constrained in practice than either its supporters or its critics expected. Attorney General Richard Hermer's legal framework — applied with a consistency that reflects genuine principle — has repeatedly constrained the government's operational options on deportation, returns, and the treatment of asylum seekers in ways that generate political cost without producing the enforcement outcomes that would justify that cost to the electorate. The result is the worst of both worlds: alienating liberal voters with the rhetoric of control while failing to deliver the control that would justify alienating them.

This is not primarily a failure of the individuals involved. It is a failure of the political system to have resolved, at any point in the preceding two decades, the fundamental tension between Britain's international legal obligations, its domestic political preferences on immigration levels, and its practical capacity to process, accommodate, and integrate arrivals at scale. The question was avoided by successive governments. It has not become easier to answer.

The entitlement question

The welfare and benefits debate in Britain has acquired a particular toxicity because it conflates several genuinely distinct phenomena: the legitimate needs of people disabled by illness or injury, the consequences of an economy that has systematically failed to create secure work at liveable wages in large parts of the country, the effects of a housing market that has transferred enormous wealth to existing property owners while pricing younger generations out, and — genuinely, if on a smaller scale than political rhetoric implies — the existence of people who have adapted rationally to a system that rewards inactivity more generously than entry-level employment.

The 'entitlement culture' critique is not wrong that such adaptation exists. It is wrong in its diagnosis of the cause — which is systemic rather than characterological — and in its prescription, which typically involves reducing benefits without addressing the structural conditions that make benefits rational for those who claim them. More relevant to this paper's argument is the political function the entitlement debate serves: it provides a domestically comprehensible, emotionally resonant explanation for Britain's economic difficulties that displaces attention from the structural causes — deindustrialisation, financialisation, Brexit's GDP drag, systematic underinvestment in productivity and technology — which are harder to understand and harder to blame on identifiable domestic actors.

In an AI century, the entitlement debate will become structurally more acute. AI-driven automation will displace employment at a rate and in categories — including white-collar, professional, and administrative work — that previous waves of automation did not reach. A political system that has spent two decades arguing about benefits without resolving the underlying conditions that generate dependency will be particularly ill-equipped to manage the social disruption of the next wave. The institutions — civil service, political parties, welfare bureaucracy — whose reform is prerequisite to managing that disruption are the same ones whose dysfunction this paper has been documenting.

Part V: Britain and the AI Century — The Collision

What Mythos reveals about British vulnerability

Project Mythos, as discussed in the companion briefing paper, is a proof of concept for AI-enabled vulnerability discovery at a scale and speed that human security processes cannot match. Its restricted release to a group of roughly 40 Western technology firms — not including most governments, central banks, or national cyber agencies — means that the entities best positioned to patch the vulnerabilities it identifies are private corporations, while the entities most dependent on legacy infrastructure and least equipped to implement rapid remediation are public institutions.

In the British context, this asymmetry is acute. The UK's critical national infrastructure — power, water, transport, financial settlement, health — runs substantially on legacy systems. The NHS, which processes data on the majority of the British population and whose operational failure directly costs lives, has been the subject of repeated major cyber attacks, most notoriously the WannaCry attack of 2017, which exploited vulnerabilities in systems running Windows XP — an operating system that had been out of mainstream support for three years. That attack was not a sophisticated nation-state operation. It was a commodity ransomware campaign that succeeded because NHS trusts had not applied available patches to outdated systems. The structural conditions that allowed WannaCry to succeed have not been fundamentally changed.

A Mythos-level capability deployed against UK critical infrastructure — whether by a hostile state, a criminal organisation, or a non-state actor that has acquired equivalent capability — would find a target environment characterised by aging systems, inadequate patching discipline, a civil service without the technical capacity to manage the response, and a political leadership without the strategic literacy to have prepared for the scenario. The NCSC's warning that the UK could be targeted 'at scale' in the event of conflict is a description of a risk that already exists and is not matched by an equivalent capacity for defence.

The hyperscaler dependency and British sovereignty

Britain's digital infrastructure is substantially dependent on American hyperscalers. Government cloud services run primarily on AWS, Azure, and Google Cloud. The communications infrastructure on which both the public sector and the economy depend uses software and hardware provided by companies incorporated in the United States and, in some cases, subject to American legal jurisdiction. The implications for British digital sovereignty are significant and have not been seriously debated in Parliament.

The Five Eyes intelligence relationship provides some protection — American hyperscalers are unlikely to be weaponised directly against British institutions. But the dependency creates vulnerabilities that go beyond the bilateral relationship. American hyperscaler systems are themselves targets for Russian, Chinese, and Iranian state actors. A successful penetration of Microsoft's government cloud infrastructure, of the kind that the Solar Winds attack approximated, is simultaneously a penetration of every government that runs its systems on that infrastructure. Britain's strategic exposure is therefore not only to direct attacks but to the cascading consequences of attacks on the private infrastructure on which it depends.

The British government's response to this dependency — pausing its Digital Sourcing Strategy, closing its own AI pilots because commercial alternatives outperform them, planning to cut civil service technical capacity — moves in exactly the wrong direction. Each step that increases dependency on private hyperscaler infrastructure and reduces public sector technical capability is a step that deepens the sovereignty deficit and widens the vulnerability window.

The leadership literacy gap

The most fundamental problem — the one that underlies all the others — is that Britain's political leadership does not understand the world it is governing. This is not a partisan observation. It applies with equal force to all the parties now contesting power. It is a structural consequence of how British politicians are recruited, trained, and selected.

The pathway to political leadership in Britain runs overwhelmingly through law, journalism, political research, and the arts. It does not, with rare exceptions, run through engineering, computer science, mathematics, or any of the technical disciplines that shape the world the political class is attempting to govern. The result is a legislature and an executive whose collective technical literacy is, by any serious measure, inadequate to the decisions they are being asked to make about AI procurement, cyber defence, critical infrastructure resilience, and digital sovereignty.

This is not simply about being able to read a spreadsheet or understand a briefing note. It is about having sufficient conceptual depth to ask the right questions, to identify when expert advice is serving institutional interests rather than national ones, to recognise the strategic significance of decisions that are presented as technical or administrative, and to build the relationships with genuine experts that would allow policy to be informed by technical reality rather than vendor marketing.

The Mandelson security clearance episode is one data point. The closure of the government's own AI pilots in favour of commercial alternatives — without any apparent strategic assessment of what that dependency means — is another. The handling of the Russia Report, delayed and redacted for political reasons, is a third. The pattern is consistent: decisions of major strategic significance are made for reasons of immediate political convenience by people who do not have the conceptual framework to understand their long-term implications.

A political class that cannot distinguish between an AI pilot and a procurement decision, between a security vetting recommendation and a political inconvenience, between a cyber incident and an act of war, is not equipped to govern in the AI century. The problem is not intelligence. It is literacy — a kind that the British political system does not currently produce or reward.

Part VI: The Defence Dimension — Decline, Transformation, and the Leadership Seesaw

The arithmetic of decline

The starting point must be the numbers, because they are more striking than most public commentary acknowledges. The British Army currently fields approximately 73,000 regular soldiers — a force that has shrunk by roughly a third since 2000, and by more than half since the Cold War peak of the 1980s. The RAF has contracted even more sharply, to around a third of its 1985 size. The Royal Navy, once the world's pre-eminent maritime power, fields ten submarines and seventy vessels in total. The full-time trained strength of all three services is around 130,000 — below target across every branch, with the RAF running 13% short of its establishment, the Royal Navy 8% short, and outflow continuing to exceed intake on a trained basis.

These numbers acquire their true significance when set against the command structure that presides over them. As of July 2025, the British Army had 211 General Officers commanding 73,000 regulars. The United States Army has 219 Generals — commanding a force of 480,000. Britain therefore has, in proportional terms, approximately five times the general officer density of the most bureaucratically top-heavy large military organisation on earth. The Royal Navy fields 134 Flag Officers for a fleet of roughly seventy vessels — approximately three admirals for every warship. The RAF has 126 Air Officers of Air Rank for a service with fewer fast jets than flag officers.

This is not simply an administrative curiosity. It is a structural diagnosis. An organisation with this ratio of senior leadership to operational capacity is not configured for warfighting. It is configured for management, administration, procurement oversight, NATO committee representation, inter-departmental liaison, and the elaborate ceremonial and institutional functions that the British military has accumulated over centuries. Some of those functions are necessary. None of them constitutes fighting power. The question — which the current reform agenda has not squarely answered — is whether the organisation can be rebalanced toward capability without dismantling the institutional culture and relationships that give it cohesion.

Britain has nearly as many Army generals as the United States — and one seventh of the soldiers. This is not a command structure. It is a bureaucracy that has accumulated military rank.

The reasons for decline: a compounding failure

The decline of British military capability is not the product of a single decision or a single government. It is the accumulated consequence of a series of choices, each defensible in isolation, that together have produced a force that is, by the frank assessment of RUSI's senior land warfare analyst, capable of fielding 'half of one heavy brigade's worth of equipment, split between two' — far short of the warfighting division that was the stated ambition of the 2015 Strategic Defence and Security Review.

The first and most persistent cause is the peace dividend that was never honestly reconciled with Britain's strategic ambitions. After the Cold War, successive governments cut defence spending while retaining commitments — to NATO, to the Five Eyes, to nuclear deterrence, to expeditionary capability, to global maritime presence — that were calibrated for a larger force. The result was a hollowing out: the capability statements remained, the resources to fulfil them contracted. Each Strategic Defence Review identified the gap; none produced the political will to close it honestly, either by funding the commitments or by reducing them to match the funding.

The second cause is the Iraq and Afghanistan legacy. Two decades of counterinsurgency operations optimised the Army for a specific kind of warfare — light, infantry-heavy, culturally complex, conducted at relatively low tempo against non-peer adversaries — while starving conventional warfighting capability of investment, training time, and doctrinal development. Heavy armour rusted, artillery shrank, combined arms integration atrophied. The skills and equipment needed to fight a peer adversary in a high-intensity, high-attrition environment were not maintained. Ukraine demonstrated, brutally and publicly, exactly what that kind of war demands and how little of it Britain's land forces could currently sustain.

The third cause is procurement dysfunction. The British defence procurement system is, by the consistent assessment of the National Audit Office, the Defence Select Committee, and successive Chiefs of the Defence Staff, among the most expensive and least effective in the developed world. Ajax, the armoured vehicle programme, spent over £5 billion and produced a vehicle that was for years too noisy and vibration-prone for crews to operate safely. Watchkeeper, the tactical surveillance drone, took sixteen years from contract to operational service. The Type 45 destroyer, one of the world's most capable anti-air warfare ships, was delivered with propulsion problems that rendered it unreliable in warm-water operations. These are not isolated failures. They are symptoms of a procurement culture that prioritises industrial policy, jobs, and ministerial announcements over military utility and delivery.

The fourth cause — directly connected to the themes of this paper — is the failure to invest in the digital and cyber dimensions of military capability at the speed the threat environment demands. The UK has endured over 90,000 cyber attacks from hostile actors in the last two years alone. Its military networks, like its civilian critical infrastructure, run substantially on legacy systems. The investment in AI-enabled military capability — while now accelerating — begins from a position of significant deficit relative to the threat.

The defence versus benefits false choice

The political framing of defence spending as a choice between guns and butter — between defence and benefits, between military capability and public services — is one of the most persistent and damaging distortions in British political discourse. It is not a false choice because defence spending is costless. It is false because it presents a structural investment decision as a simple zero-sum trade-off, obscuring both the true cost of inadequate defence and the true scale of the fiscal space available for different priorities.

The numbers warrant examination. Britain currently spends approximately 2.3% of GDP on defence — below the NATO average, and, crucially, almost certainly below the 1.5–1.7% that represents genuine warfighting capability once pension accounting, intelligence budgets, and civil servant costs are stripped out. NATO members have now agreed a new target of 3.5% of GDP on core defence by 2035, with a further 1.5% on defence-related investment, totalling 5%. Meeting the core 3.5% target from Britain's current position would require additional annual spending of roughly £25–30 billion. That is significant. It is not, in the context of a total public spending envelope of over £1.2 trillion, existential.

What makes the choice feel impossible is not the absolute cost but the political economy of reallocation. Britain spends approximately £280 billion per year on social protection — welfare, pensions, housing benefit, and related transfers. The political parties that have been most aggressive in cutting defence have been most reluctant to reform the social protection system in ways that would generate fiscal headroom. The parties most willing to spend on defence have been most hostile to the tax increases that would fund it without cuts elsewhere. The result is a political system that has no credible mechanism for making the trade-offs honestly, and therefore does not make them at all.

The deeper problem is that the choice between defence and benefits is, in strategic terms, a choice between present consumption and future survival. A welfare system that produces a population healthy, educated, and economically secure enough to staff a military, sustain an industrial base, and maintain social cohesion under the pressure of a serious security threat is not in competition with defence — it is a precondition for it. The countries that have most successfully combined strong welfare provision with serious military capability — the Nordic states, the Netherlands, Germany since 2022 — have done so by treating both as non-negotiable and finding the fiscal room through growth, efficiency, and honest taxation. Britain has, for thirty years, treated both as negotiable and found neither adequately funded as a result.

The choice between defence and benefits is not, in the end, a choice between different public goods. It is a choice between a state that is resilient and one that is vulnerable — and the countries that have made that choice most clearly have not had to sacrifice either.

The changing nature of defence: what warfare now demands

The war in Ukraine has provided the most vivid and violent demonstration of what twenty-first-century peer conflict actually requires — and the gap between that requirement and Britain's current posture is considerable. Ukraine has produced drones at a rate of 4.5 million per year. The European Commission estimates that defending a country the size of Lithuania in a wider war with Russia would require three million drones annually. Britain's current drone production and acquisition programme, however improved, is not calibrated to this scale. The Strategic Defence Review's '20/40/40' model — 20% traditional platforms, 40% expendable autonomous systems, 40% reusable AI-enabled assets — is the right intellectual framework. The question, as RUSI analysts have noted, is whether the budget and the procurement system can deliver it.

The tempo of AI-enabled warfare poses a specific challenge that goes beyond equipment. Ukraine's Operation Spiderweb in June 2025 — over 100 drones striking multiple Russian air bases simultaneously — was followed within three months by Russian integration of AI-powered drones into its own operations. The cycle from observation to doctrine to deployment has compressed from years to months. A military whose procurement cycle runs to decades, whose doctrine development is institutional and slow, and whose senior leadership is oriented toward management rather than operational innovation, is structurally mis-matched to this tempo.

The British Army's ASGARD targeting system — which aims to reduce the kill chain from target detection to strike decision to under a minute — is a genuine step toward the speed of AI-enabled warfare. The Atlantic Bastion programme for autonomous submarine-hunting in the North Atlantic is conceptually sound and urgently necessary given Russian undersea activity around British and NATO cable infrastructure. The DragonFire laser weapon system, due to be fitted to Royal Navy destroyers from 2027, represents genuine technological leadership. These are real achievements. They do not, taken together, constitute a transformation of Britain's warfighting posture. They are innovations within a structure that remains, in its fundamentals, configured for a world that no longer exists.

The leadership seesaw: what military command now requires

The fundamental question about British military leadership is not whether the individuals are talented — they demonstrably are, across all three services. It is whether the system selects, develops, and empowers the right kind of leadership for the AI age of warfare, and whether the institutional structure gives that leadership the room to operate effectively.

The general officer ratio problem — 211 Army generals for 73,000 soldiers — is a symptom of a deeper structural pathology: a military that has expanded its command superstructure as its fighting strength has contracted. Senior officers spend careers in procurement oversight, NATO committee work, inter-departmental liaison, and ministerial briefing. These are necessary functions. They are not, however, the functions that develop the operational judgment, technical literacy, and strategic instinct that AI-era warfare demands. The United States, for all its institutional inefficiencies, has built a cadre of senior officers who understand autonomous systems, AI-enabled decision-making, and multi-domain integration because their careers have included direct engagement with those capabilities at operational level. The British system, by contrast, tends to reward institutional navigation over operational innovation.

The specific literacy gap that matters most for the AI century is the gap between military leadership and the technological systems on which military capability now depends. A general who cannot interrogate the assumptions embedded in an AI targeting system, who cannot evaluate the security architecture of a military communications network, or who cannot assess the strategic implications of an adversary's AI capability, is not equipped to command effectively in the environment that is now arriving. This is not a criticism of individuals. It is a description of a system that has not yet built the pathways through which technical literacy becomes a prerequisite for senior command rather than an optional extra.

The corollary is that the civilian Ministry of Defence, which should provide the strategic oversight and resource management that frees military leadership to focus on operational capability, has its own version of the same problem. MoD civil servants who cannot evaluate the strategic significance of an AI procurement decision, who cannot assess the security implications of a defence digital dependency, and who cannot hold major contractors accountable for technical delivery, are not providing effective oversight. They are providing expensive administration. The combination — a military leadership oriented toward institutional management and a civilian oversight structure without technical depth — is the worst of both worlds for a country attempting to transform its defence posture in the AI era.

The model that the evidence points toward is one in which the general officer establishment is substantially reduced — not by cutting the functions those officers perform, but by asking whether those functions require general officer rank, and whether they are the right functions for an organisation with Britain's current and projected resource level. A smaller, more operationally focused senior officer corps, with genuine technical literacy embedded as a requirement rather than a bonus, and with career pathways that reward innovation and operational excellence over institutional longevity, would be a more effective instrument of national defence than the current structure. It would also free resources — both financial and human — for the autonomous systems, AI-enabled command infrastructure, and digital warfighting capability that the strategic environment now demands.

Britain does not need fewer good officers. It needs a different kind of officer — one whose career has been shaped by the technology of warfare rather than the administration of defence. The distinction matters more than it has ever mattered before.

Defence and the 'ungovernable' thesis: the closing argument

The defence picture connects to the 'ungovernable' thesis in a specific and important way. A state whose military capability has been systematically degraded over three decades, whose defence procurement system consistently fails to deliver, whose senior command structure is optimised for administration rather than warfighting, whose political leadership lacks the literacy to make sound defence decisions, and whose critical infrastructure has been penetrated by hostile states, is not merely poorly governed. It is strategically exposed in ways that constrain its freedom of action and ultimately its sovereignty.

Britain's exposure is not yet existential — the nuclear deterrent, the intelligence relationships, the quality of individual military personnel and units, and the residual capability of GCHQ and the intelligence services provide real strategic weight. But the trajectory is clear, and the AI century is accelerating it. Every year that the defence transformation is delayed, every procurement programme that fails to deliver, every general officer post that absorbs budget without producing fighting power, every political decision that treats defence as a cost to be minimised rather than a capability to be built, widens the gap between Britain's strategic commitments and its capacity to fulfil them.

The ungovernable thesis, applied to defence, is this: a political system incapable of making honest choices about resource allocation, populated by leaders without the technical literacy to understand the choices before them, operating through institutions configured for a different era, will not — in the absence of a genuine crisis — generate the political will to make the investments and reforms that Britain's defence posture requires. That is not ungovernability in the absolute sense. It is strategic drift. In an AI century, strategic drift is not a stable condition. It is a deteriorating one.

Part VII: Is Britain Ungovernable — And What Would Change That?

The honest answer

Britain is not ungovernable in the absolute sense. Its institutions function. Its courts remain independent. Its armed forces remain professional and well-led. Its intelligence services, despite the political pressures documented above, retain significant capability. The NCSC is among the most respected cyber security agencies in the world. The country continues to produce world-class science, engineering, and creative work. The diagnosis of total ungovernability is, as the Institute for Government has argued, a dereliction of responsibility — it exculpates those who govern badly by attributing their failures to systemic impossibility.

But the conditions that would produce genuine ungovernability are assembling. The combination of a political system in decomposition, a civil service without the technical capacity the moment demands, a decade of accumulated hostile state penetration whose effects are unknown but likely significant, critical infrastructure that is inadequately resilient, a leadership class without the strategic literacy the AI century requires, and a social contract under strain from unresolved questions of migration and entitlement — this combination does not produce a state that cannot be governed. It produces a state that is extremely vulnerable to the kind of compounding crisis that the AI century is likely to generate.

What would actually change it

The changes required are not mysterious. They are simply politically difficult in the current environment, which is precisely why they have not been made.

Electoral reform that produces a parliament genuinely representative of the electorate would not solve the substantive policy problems but would restore the legitimacy that is the prerequisite for addressing them. A first-past-the-post system that delivers Reform majorities on 25% of the vote will not produce stable government or legitimate authority.

A fundamental reconfiguration of the civil service — not reduction but transformation — that recruits, trains, retains, and promotes technical expertise at the same level as generalist administration, and that builds genuine in-house AI and cyber capability rather than outsourcing it to the private sector, is the prerequisite for effective governance in the AI century.

A serious, public account of the accumulated damage done by hostile state cyber and intelligence activity — equivalent to the kind of reckoning that countries like Estonia underwent after Russian cyber attacks in 2007 and used as a catalyst for genuine transformation — is necessary to build the political will for the investment in resilience that the threat requires.

Most fundamentally, a political culture that rewards strategic literacy — that asks not just 'who do you represent' and 'what do you believe' but 'do you understand the world you are proposing to govern' — would change the quality of the people making these decisions. That is a longer-term transformation. It begins with being honest about the scale of the problem.

The Mythos question as a test case

Project Mythos is, in this context, a test of whether Britain has the institutional capacity to act in its own strategic interest in the AI century. The UK is not in the Glasswing group of 40 institutions with access to the tool. Its critical infrastructure contains the legacy vulnerabilities that Mythos is designed to find. Its cyber defence posture is improving but remains structurally inadequate. Its political leadership has not, in public, engaged with the strategic significance of what Mythos represents.

A government with the strategic literacy this paper argues is necessary would be asking three questions. First, what is in Mythos or its successors that applies to UK critical infrastructure, and how do we get access — or build equivalent capability? Second, what does the existence of such a tool tell us about the security of our legacy systems, and what is the remediation plan? Third, what does the restricted-access model — private access for commercial partners, nothing for most governments — tell us about the adequacy of existing multilateral governance frameworks, and what is Britain's position on building alternatives?

These questions are not being asked publicly. Their absence from British political discourse is a symptom of exactly the literacy deficit this paper has attempted to diagnose. The world is not waiting for that deficit to be addressed. The AI century is already here.

END OF ANALYSIS

This paper should be read in conjunction with 'Resilience in the AI Century: Hyperscalers, Strategic Dependency, and the Case for a New Security Doctrine' (May 2026).