Humanity's Last Defence Review?

Testing OpenAI's new Deep Research model in ChatGPT, by asking it to undertake the UK Defence Review

There has a been a lot of commentary on DeepResearch, the latest model from OpenAI. Tyler Cowan writes on his Marginal Revolution blog that:

“I have had it write a number of ten-page papers for me, each of them outstanding. I think of the quality as comparable to having a good PhD-level research assistant, and sending that person away with a task for a week or two, or maybe more.

Except Deep Research does the work in five or six minutes. And it does not seem to make errors, due to the quality of the embedded o3 model.

It seems it can cover just about any topic?”

What is DeepResearch you ask? OpenAI tells us it is ‘…a new agentic capability that conducts multi-step research on the internet for complex tasks.’

I thought I’d have it write the UK Defence Review, an experiment. I share because I think it is eye-opening. If you want the conclusions, skip to the end, but I encourage you to read it.

The first version I produced drew mainly on the Whitehall think thanks and existing UK Government documents. This was partly because my prompt asked for sobriety. It read to me like precisely the kind of thing Government usually produces - bland, prioritising politics and compromise over confronting issues and coherence.

As the prompt shows, I had it redo it twice more. First to re-write around a clearer theory of victory (which it didn’t do very well, but I think that is partly because in the UK sources it was reading, no-one has really advanced one).

So in the second addition to the prompts, I asked it consider my own writings, those of Ken Payne, Mike Horowitz, Edward Stringer, Christian Brose, Raj Shah and Christopher Kirchhoff, Dominic Cummings, Defence and Security reforms in Scandinavian countries, the Baltics and Japan. I picked those who are in one way or another quite radical, which also I think makes it a better read for the experiment and publication here.

I don’t doubt many of you would have prompted quite differently, and I cringe a little bit about asking it to look at my work. But on the other hand if it is on this blog it might be odder if I didn’t shape it to be like something I would (largely) agree with.

This took 6 minutes for Deep Research to produce. It took me significantly longer to cut/paste and fix the formatting. It is almost certain that any formatting or hyperlink errors in the below are my mistakes, and weren’t in the original.

Here’s the output:

---

UK Strategic Defence Review 2025: A Bold and Unflinching Overhaul

Executive Summary

Britain’s 2025 Strategic Defence Review calls for radical change to ensure the UK armed forces can fight and win in tomorrow’s battles. This review makes bold decisions to overhaul force structure, embrace emerging technologies, and ruthlessly cut waste. Key imperatives include:

• Pivot to a Kill-Chain Warfare Model: Reorganize around fast, networked kill chains – the ability to find, target, and strike enemies with unprecedented speed​

  (ndupress.ndu.edu). The side with the fastest, most effective kill chain will dominate in modern war​ (ndupress.ndu.edu). This means linking every sensor to shooters via resilient networks, powered by AI, to compress decision cycles.

• Aggressive Tech Integration (AI, Drones, Cyber, Space): Make AI and automation the backbone of UK defense. AI-enabled decision support and autonomous systems can outpace human reaction, giving a decisive edge​ (Bassett-Cross & Dear, 2024). From swarming drones to space-based assets, the UK must harness technologies that accelerate the kill chain and multiply lethality (ndupress.ndu.edu). We will not be a mere observer in the AI revolution – we will lead it.

• Force Structure Overhaul: Build a leaner, highly lethal, networked force. We will field many more expendable, smart weapons (drones, loitering munitions, autonomous vehicles) and fewer expensive legacy platforms. This “Mosaic” or “Moneyball” approach returns mass and resilience to the battlefield by overwhelming adversaries with swarms of low-cost targets​ (hoover.org). China or Russia would face thousands of UK autonomous systems instead of a handful of inflexible units​ (hoover.org). Every unit and domain (land, sea, air, cyber, space) will be linked and capable of both sensing and shooting.

• Hard Cuts to Legacy Systems: We will divest from systems that do not survive or deliver in future warfare. If a platform cannot contribute to a rapid kill chain or is too easily picked off by modern precision weapons, it will be reduced or eliminated. For example, heavy armored vehicles and manned artillery that are vulnerable to cheap drones and smart missiles have a diminished role​ (scsp.ai). Expensive programs that squander billions without delivering capability (e.g. past procurement “farces”​(ukdefencejournal.org.uk) will be halted. Every pound saved will be reinvested into high-impact capabilities.

• Reshape Industry and Procurement: Break the cycle of slow, over-budget acquisition. We will create a fast-track “Unit X” style innovation pipeline to inject cutting-edge commercial tech into the Armed Forces​ (Unit-X). Bureaucratic red tape will be slashed – speed and outcomes will define procurement success, not process. We’ll partner with the UK tech sector (AI labs, startups, advanced manufacturers) just as Ukraine leveraged SpaceX for resilient comms in war​ (scsp.ai). A new “British DARPA” ethos will fund high-risk, high-reward projects and field prototypes within months, not years.

• Disciplined Execution & Timeline: This transformation will happen on a wartime footing. Clear deadlines will drive change – no more decades-long programs that deliver too late​ (defensenews.com). Leadership will be held accountable for meeting milestones in force restructuring, tech deployment, and savings reinvestment. The pace of global threats (from China’s rapid military-tech advances​ (ndupress.ndu.edu) to Russia’s aggression) leaves no room for delay or half-measures.

Bottom Line: The 2025 Review delivers an unflinching plan to make Britain’s military fit for the future. It sacrifices the comfortable old ways in favor of bold action and innovation. By embracing new warfare concepts and slashing inertia, the UK will ensure it can deter, fight, and win against any adversary in the coming decades. The following sections detail the theory of victory and the specific changes required in force structure, technology, industry, and execution to achieve that end state.

Theory of Victory in the Future Battlespace

How will Britain deter and win wars in the 2030s and beyond? Our theory of victory centers on outpacing and outsmarting the enemy through superior networks, information, and agility. In an age of high-tech confrontation, victory will not come from sheer volume of troops or the longest-range gun – it will come from turning inside the enemy’s decision loop and denying them any easy gains. Key elements of this theory include:

• Kill-Chain Dominance: We will achieve decision superiority by executing the kill chain (find-fix-finish) faster than any opponent​ (ndupress.ndu.edu). If we can see and strike the enemy before they can react, we win. This means integrating intelligence, surveillance, and reconnaissance (ISR) with shooters via AI-enabled command systems. For example, UK forces will fuse satellite imagery, cyber intelligence, and forward sensors in real time to direct precision strikes within minutes. In modern war games, forces that master this kind of rapid targeting “have nearly perfect records” of victory, whereas those who lag lose badly​(ndupress.ndu.edu). Our theory of victory is to always be the side with the fastest kill chain.

• Deterrence by Denial and Rapid Punishment: Britain will deny adversaries their objectives from Day 1 of a conflict. Borrowing from the Baltic states’ resolve, we will make any attempted attack immediately costly and likely to fail. The Baltic and Nordic countries have shown the importance of defending the entire national territory from the outset, rather than ceding ground​ (osw.waw.pl). Likewise, the UK will field forces (including resilient local defenses, long-range strike, and cyber units) that can blunt and punish an aggressor immediately. By having the capability to “find and finish” enemy units swiftly via networked fires, we ensure any aggressor knows they cannot achieve a quick, low-cost win​ (scsp.ai). This deters war by raising the risks to the initiator.

• Multi-Domain Integration: Victory will rely on seamlessly combining effects across all domains – land, air, sea, cyber, and space. The UK’s theory of victory leverages synchronized operations: e.g. blinding enemy sensors via cyber or electronic attack, while precision missiles from air and sea strike critical targets, and autonomous drones saturate the battlespace. Each domain reinforces the others in a web of capabilities that an adversary cannot easily unravel. This system-of-systems approach mirrors the way Ukraine has fought off a larger foe by networking drones with artillery and space-based comms​ (ndupress.ndu.edu; scsp.ai). By attacking along every axis (physical and digital) faster than the enemy can respond, we achieve an “offset” – rendering their strengths (like massed tanks or ships) ineffective.

• Continuous Innovation & Adaptation: A core pillar of our theory is that the first side to innovate wins. We face a rapidly evolving threat environment – notably the AI revolution in warfare. Our strategy is to lead technological change, not follow. As one expert put it, you need a “theory of winning” that anticipates the future and builds for it, rather than waiting for the future to arrive​ (Dear, Theory of Victory, 2024).

• This means the UK will constantly wargame and experiment with new concepts (from swarming munitions to AI battle management) and quickly adopt what works. We recognize that standing still means falling behind in the face of adversaries (like China) who are innovating at speed​ (politico.eu). Thus, adaptation and learning are built into our operations – our doctrine will evolve as fast as technology does.

• Alliance Leverage and Niche Strengths: Finally, Britain’s theory of victory acknowledges we fight as part of alliances (NATO and partnerships in the Indo-Pacific). We will focus on niche advantages that complement allies – for instance, excelling in certain tech (cryptography, cyber offense, undersea warfare) or unique force packages – to punch above our weight. By contributing disproportionately valuable capabilities to coalitions, we help tip the overall balance in our favor. At the same time, we will not rely on others for critical gaps; the UK must be able to act independently in a pinch. That requires sustaining capabilities like a minimum nuclear deterrent and strategic lift, while modernizing conventional forces to be self-sufficient for homeland defense.

In summary, our theory of victory is about brain over brawn: using intelligence, speed, and innovation to defeat threats. We deter by convincing any adversary that the UK can out-target, out-think, and out-last them in conflict – making aggression futile. Should deterrence fail, we will fight as a nimble, tech-empowered David against any Goliath, confident that precision and networking can defeat brute force. This theory underpins the force design and investments outlined in the next section.

Force Structure Overhaul: Building the Future Force

The UK Armed Forces will undergo a radical force structure transformation to implement this vision. We are moving from a legacy force built for industrial-age warfare to a digital-age force optimized for distributed, network-centric operations. This overhaul prioritizes quality and connectivity over sheer size, and agility over heavy mass. Key changes include:

• Smaller, Networked Combat Units: We will break our forces into agile, networked units that can operate semi-independently yet share data instantaneously. Battalion-sized formations will be augmented (or even replaced) by “unit cells” equipped with their own drones, sensors, and long-range fires, all linked via secure communications. Instead of large brigades needing massive logistics, many smaller units will swarm the adversary from multiple directions, complicating their targeting. This is inspired by concepts like the US “Mosaic Warfare” and the Ukrainian model of dispersed resistance, which show that many small, smart units can outperform a few big ones. Importantly, every unit will be capable of plugging into a larger combat cloud – a resilient network that shares targeting data and intelligence in real time.

• Return of Mass through Autonomy: Paradoxically, we will increase “mass” (number of platforms and weapons) while downsizing manpower. The answer is autonomous systems in bulk. The UK will deploy swarms of low-cost drones, uncrewed vehicles, and smart munitions to regain numerical advantages that were lost when we relied only on a few high-value assets. A “Moneyball Military” approach aims to field quantity with quality – inexpensive, attritable systems that are networked together​ (hoover.org). In a conflict, dozens of autonomous drones might accompany each squad or ship, overwhelming enemy defenses by sheer numbers. Because each drone or robot is relatively cheap, losing some is acceptable – they impose disproportionate costs on the enemy by forcing them to waste expensive missiles or expose themselves to counter-fire​ (hoover.org). This strategy forces the enemy into a dilemma: if they don’t shoot, our swarms strike them; if they do shoot, they reveal their positions and exhaust their munitions on less valuable targets​ (hoover.org; hoover.org). We saw glimpses of this in Nagorno-Karabakh and Ukraine, where drone swarms decimated armour and artillery​ (scsp.ai). The UK will lead in fielding thousands of autonomous systems, from air-launched drones to uncrewed submarines, restoring combat mass and complexity in our favor.

• Cutting-Edge “Kill Web” Architecture: The new force will be designed around a “kill web” – a web of sensors and shooters – rather than around traditional service silos. In practice, this means a Royal Air Force fighter might cue a British Army long-range missile launcher to fire, based on Navy drone surveillance data. All components act as part of a unified kill chain network. To enable this, we’ll invest heavily in C4ISR (Command, Control, Communications, Computers, Intelligence, Surveillance, Reconnaissance) that is secure and jam-resistant. Every platform – whether a tank, ship, aircraft or satellite – will be a node that can send and receive targeting data. The principle is “Every sensor, every shooter” – if one unit sees a target, any unit can kill it. This will require common data standards and intense training in joint operations. The payoff is enormous: a resilient kill web means even if one node is destroyed, the network fights on. It mirrors how Russia’s enemies demonstrated high-speed kill chains by pairing old artillery with new drone reconnaissance, achieving effects disproportionate to their resources​ (ndupress.ndu.edu). Our overhaul ensures the UK always fights as an integrated network, not separate service fiefdoms.

• Highly Lethal, Niche Capabilities: Within this new structure, we will emphasize high-lethality units armed with next-generation capabilities. For example, long-range precision fires units will be expanded – including land-based rocket forces able to strike targets hundreds of kilometers away with precision. Investments in capabilities like hypersonic missiles and advanced anti-ship missiles will give even small UK units a big punch. We will also stand up dedicated “drone swarms battalions” or squadrons – units trained specifically to deploy and manage swarms of loitering munitions and autonomous drones for reconnaissance and attack. The Air Force will incorporate loyal wingman drones to fly alongside piloted jets, multiplying their effectiveness. The Navy will introduce more uncrewed surface and underwater vehicles to extend the fleet’s reach (for example, uncrewed vessels scouting ahead of manned ships or protecting the strategic nuclear deterrent submarines). Special Forces and cyber units will also grow, focusing on high-impact niche roles like strategic sabotage, counter-space operations, and AI-enabled psychological operations. In short, every part of the force is being re-tooled to hit above its weight class through technology and smart design.

• Sustainable and Agile Logistics: A truly agile force overhaul must address logistics – often the Achilles’ heel. We will move away from ponderous supply convoys and vulnerable supply depots, towards just-in-time resupply, 3D printing at forward locations, and unmanned delivery (e.g. drones dropping supplies to front-line units). Smaller units with autonomous platforms also mean leaner support requirements (many drones are electric or simpler than crewed vehicles). The force will pre-position caches of munitions and spare parts in dispersed locations (including at sea on auxiliary ships or offshore platforms) to sustain high-tempo operations. We’re also exploring “resupply swarms” – autonomous cargo vehicles that can navigate to units under fire, meaning the fighting force can keep moving without waiting for traditional logistics. This agile sustainment model is informed by observing how Ukrainian units, using civilian quadcopter drones and local 3D-printed components, maintained their fight despite disrupted supply lines. For the UK, ensuring our kill chain and combat units remain effective in a protracted high-tech war means reinventing logistics with the same innovation we apply to combat units.

This force structure overhaul is unapologetically bold. It will remake the British military into a leaner, tech-heavy, networked force that is deadly and hard to defeat. We acknowledge that fewer soldiers or big platforms does not mean less power – on the contrary, if each remaining unit is backed by swarms of smart weapons and a superior network, our combat power and deterrence credibility will vastly increase. By doing this now, we get ahead of the curve of military transformation, rather than lagging behind and risking defeat​ (Dear, Theory of Victory, 2024).

Cuts and Reallocation: Ruthless Prioritization

To fund and empower the future force, the MoD will make tough, unsentimental cuts to programs and forces that no longer serve our strategic needs. This Review is explicit about what gets cut and why, breaking the cycle of spreading resources too thinly across legacy programs. Every cut is aimed at eliminating inefficiency or vulnerability and freeing up investment for higher-impact capabilities. Key decisions include:

• Reducing Heavy Armor and Manned Artillery: The role of heavy main battle tanks and tube artillery will be scaled back. Lessons from recent conflicts show that such armored forces, if not networked and protected, become prey for modern precision weapons and drones​ (scsp.ai). Rather than maintain a large fleet of heavy tanks, the Army will retain a smaller core of upgraded tanks for niche uses (e.g. urban breakthrough or as protected fire bases) and redirect funding to mobile anti-armor teams, precision-guided missiles, and drone support. Self-propelled artillery units that lack range or survivability against counter-battery fire will be trimmed; emphasis shifts to long-range rocket artillery (e.g. HIMARS-type systems) and loitering munitions that can strike without a large signature. These cuts recognize that a $100k smart munition can destroy a $5 million tank, so we will invest in the $100k munition rather than add more $5m targets for the enemy. The traditional heavy cavalry is thus no longer a growth area – agility and range are the new firepower.

• Mothballing or Selling Off Legacy Aircraft and Ships: We will retire platforms that are too expensive to operate and insufficiently adaptable. Some older Typhoon fighter jets, which will struggle against advanced air defenses and drones, will be phased out faster in favor of newer F-35s and unmanned combat aerial vehicles. Similarly, the Royal Navy will consider decommissioning aging frigates or minehunters that cannot be affordably modernized, reallocating their crews and budgets to frigates with advanced anti-air/anti-submarine capabilities and to uncrewed vessels. No “sacred cows” are exempt – if an asset does not contribute to the future kill chain or cannot be upgraded with networking and AI, it will be a candidate for reduction. This candor follows Dominic Cummings’ critique that MoD often “squanders billions” on legacy purchases that mainly enrich contractors​ (ukdefencejournal.org.uk). We will not repeat that mistake; platforms that don’t pull their weight will go.

• Streamlining Personnel and Commands: A futuristic force needs a high-skill, lean workforce. The review calls for cutting administrative overhead and possibly reducing personnel in certain traditional roles, while increasingpersonnel in tech-focused roles. We anticipate trimming some senior ranks and HQ staff (“tail”) to protect investment in frontline “teeth.” For example, the Army’s total manpower might slightly decrease as some infantry battalions are stood down – but simultaneously we will grow units like the Ranger Regiment (for special operations), cyber forces, and technical cadres for drone operations. The net effect shifts human resources from quantity to quality. Additionally, duplicative command structures will be merged to force tri-service integration. A slimmer command structure also means faster decision-making and less budget spent on bureaucracy. These manpower reallocations ensure we have the right people (data scientists, drone pilots, AI engineers, etc.) instead of large numbers in legacy occupations that technology can augment or replace.

• Ending Failing Procurement Programs: Perhaps most emblematic, we will not pour more money into programs that have shown they cannot deliver. The Ajax armored vehicle program, for instance, has been plagued by years of delays and technical problems – initially due in service in 2020, now pushed to 2025-2029 for full capability​ (defensenews.com). Such a timeline is unacceptable in a fast-changing threat environment. If ongoing reforms cannot quickly salvage Ajax, we will cut our losses and reallocate its £5+ billion budget to proven or emergent systems (like advanced infantry fighting drones or extra precision missiles for existing launchers). The same scrutiny applies to any big project running significantly over time or budget. Parliament and the public demand accountability: “Enough is enough” on programs that devour resources without output​ (defenceprocurementinternational.com). By cancelling or downsizing these, we free up funds for systems that can be deployed now or in the very near future.

• Defense Estate and Legacy Infrastructure: We will also look at closing or repurposing military bases and facilities that were built for a force structure of the past. Maintaining excess real estate diverts funds from combat power. Consolidation of training sites, outsourcing of some non-core services, and energy-efficient base upgrades will save money over the next decade – those savings will be funneled into combat capabilities. Similarly, stockpiles of old munitions or equipment that are obsolete will be retired (with any useful items sold to allies or used for training), reducing storage and maintenance costs.

All savings from these cuts are ring-fenced for reinvestment into the new capabilities identified in this Review – there is no “peace dividend” being pocketed elsewhere. This aggressive reallocation turns slack and waste into fuel for innovation. We recognize these choices have impacts – jobs lost in some traditional sectors, sentimental attachments to historic regiments or ships – but the overriding criterion is combat effectiveness for the threats of tomorrow. As painful as change can be, the greater pain would be sending our troops to fight with the wrong tools. This Review chooses future readiness over past habits, firmly and without apology.

Technology and Kill Chain Acceleration

Modern technology – especially AI and autonomous systems – is the center of gravity for future military power. To ensure the UK is not outpaced, we will launch an aggressive plan to integrate emerging technologies across all defense activities. This is about dramatically accelerating the kill chain: using tech to see, decide, and strike faster and more precisely than ever. Key initiatives include:

• AI-Enabled Decision Making: We will embed artificial intelligence at every level of command, from strategic planning to real-time targeting. Advanced AI decision-support tools can analyze vast data (satellite feeds, intelligence reports, sensor readings) far quicker than human staff, cueing commanders to threats and options in seconds. For instance, AI systems will help identify targets in live drone feeds or predict enemy moves by analyzing patterns – tasks that took analysts hours can be done in near-real time. The goal is to compress the OODA loop (Observe–Orient–Decide–Act) to a degree our adversaries cannot match. Already, algorithms can beat humans in complex strategy games and handle massive information streams; applied to warfare, even today’s AI (not to mention tomorrow’s) would be “completely transformative” if fielded at scale​ (Bassett-Cross & Dear, 2024). As one UK defence advisor put it, if we adapted current AI breakthroughs for military use, it would revolutionize operations – yet nothing as advanced is currently deployed in UK defence​ (Bassett-Cross & Dear, 2024). We will change that at speed.

• Autonomous Weapons and Platforms: Embracing autonomy is essential to winning the kill-chain competition​ (ndupress.ndu.edu). The review mandates rapid development and acquisition of autonomous weapons – from loitering munitions that can hunt targets independently, to uncrewed combat aircraft, robotic ground vehicles, and AI-driven naval vessels. These systems react faster than human-operated ones and can be risked in dangerous roles without fear of casualties. The ethical standard we will apply is pragmatic: an autonomous system doesn’t have to be perfect, it just needs to make better decisions under combat stress than a human would​ (ndupress.ndu.edu). Many autonomous platforms already meet that bar in specific tasks (e.g. automated point-defense guns can shoot down incoming missiles faster than human operators). We will deploy autonomous wingmen drones alongside fighter jets, autonomous sentry towers to guard perimeters, and swarm drones that can coordinate attacks on enemy air defenses. By 2030, a significant share of UK firepower will be delivered by uncrewed systems – dramatically speeding up the kill chain and saturating the enemy with threats.

• Long-Range Precision Fires & Hypersonics: To strike first, we must strike farther and faster. The UK will invest heavily in long-range precision fires, including land-attack cruise and ballistic missiles, and will join allies in developing hypersonic missiles that can reach targets in minutes. This extends our kill chain across theater distances. Having the ability to hit an adversary’s critical assets (like bases or ships) from hundreds or thousands of kilometers away complicates their plans and keeps them off-balance. Japan’s recent decision to acquire long-range strike weapons (breaking a historical norm) shows the importance of this capability in deterring aggression​ (rusi.org). We will ensure the RAF and Royal Navy have inventories of conventional precision missiles that can be fired from air, sea, or land, giving Britain a global strike reach. In parallel, we are developing directed-energy weapons (lasers and microwave systems) to counter the enemy’s precision munitions and drones – adding a layer to our kill chain that intercepts incoming threats at the speed of light.

• Resilient Networks & Battle Management: All these tech advantages hinge on robust, secure connectivity – our “nervous system” in war. We are going to harden and upgrade our networks using new communication satellites, encrypted tactical datalinks, and backup methods (like mesh networks formed by drones or balloons). The Ukraine war underscored this: having services like SpaceX’s Starlink ensured Ukrainian forces stayed connected even under heavy electronic attack​ (scsp.ai). We will similarly partner with commercial providers to augment military satcom and internet, ensuring our forces never go dark. Moreover, we’re fielding advanced battle management software (often AI-driven) to orchestrate forces in real time. Picture an AI “chief of staff” that helps a commander rapidly assign targets to the optimal shooter – that is the kind of system we are deploying to knit together our kill webs. Cyber defense and electronic warfare improvements also come under this umbrella: our networks must resist jamming and hacking. In exercises, we will practice “fighting fractured” – assuming networks are contested – to refine the tech and tactics for staying linked under fire. A resilient networked force can coordinate kill chains even in chaos, whereas an adversary with disrupted comms will be paralyzed.

• Military Application of Emerging Tech (Quantum, Biotech, etc.): Beyond AI and autonomy, we’re scanning the horizon for any technology that could provide a war-winning edge. For example, quantum computing and sensing could revolutionize intelligence (cracking codes, detecting submarines) – we’re investing in R&D now so we aren’t surprised later. Biotechnology advances might enhance soldier performance or create new materials for defense; we’ll watch those closely too. The key is an open mindset: as one defense innovator stressed, you “don’t wait for the future to arrive – imagine it and work backwards to build what you need”​ (Dear, Theory of Victory, 2024). In that spirit, we’ve established “Unit X” style incubators and sandboxes to quickly test cutting-edge ideas from labs or startups. Whether it’s drone swarms guided by neural networks or augmented reality helmets for troops, if it can sharpen our kill chain or make our force more lethal and survivable, we will experiment with it immediately. The MoD will take calculated risks on prototypes and pilot programs, knowing some will fail – but those that succeed could be game-changers.

All these efforts combine to accelerate the kill chain – the time from detecting a threat to eliminating it. We’re striving for what the U.S. calls “decision advantage” and “dynamic strike” capabilities, scaled to a nation of Britain’s means. Crucially, this isn’t just about buying gadgets; it’s about integrating them into doctrine and training. We will train AI-assisted operations and human-machine teaming as a norm. Troops will learn to trust and effectively employ autonomous systems. Commanders will learn to delegate certain tasks to AI and focus on the bigger picture. By vigorously pursuing these tech advances, the UK will not only keep up with but in some areas outpace the Pentagon and PLA in military innovation. Indeed, British defence thinkers have warned that if we don’t harness AI and emerging tech, we risk becoming an “AI colony” of those who do​ (politico.eu). This Review ensures Britain remains a leader, not a follower, in wielding the technologies that are reshaping warfare.

Industrial and Procurement Revolution

Achieving all the above requires a fundamental change in how we develop and acquire military capability. The traditional UK defence procurement system is too slow, too risk-averse, and too insular to deliver the needed innovation on time​ (ndupress.ndu.edu). We are therefore unleashing an Industrial and Procurement Revolution to rapidly translate money and ideas into fielded capability. Key components of this revolution are:

• Embrace a “Unit X” Model: The MoD will establish a new high-level Defence Technology Accelerator – effectively a UK version of the U.S. Defense Innovation Unit (codenamed “Unit X”). Its mission: bridge the gap between the military and the tech sector​ (Shah & Kirchhoff, 2024).

  This organisation will be empowered to bypass normal bureaucracy and work with non-traditional suppliers (startups, academia, big tech firms) to solve military problems with cutting-edge tech. Like a venture capitalist, it can rapidly fund and trial solutions, focusing on speed and outcomes over process. Unit X in the US was able to pull technologies like AI drones and microsatellites into use much faster than traditional programs​ (Shah & Kirchhoff, 2024). We will do the same, making the UK a welcoming partner for innovators. This also feeds our local tech industry – rather than relying only on a few big defense primes, we’ll tap the creativity of Britain’s wider science and technology base. In an era when China practices “Military-Civil Fusion” – requiring all companies to support the PLA​ (Shah & Kirchhoff, 2024) – it’s imperative we mobilize our private sector genius for national security as well. The old walls between “Silicon Roundabout” (UK’s tech hub) and Whitehall must come down.

• Speed Over Perfection: A key cultural shift in procurement will be adopting a “fast enough to fail, fast enough to fix” mentality. Rather than spend 15 years trying to develop the perfect platform (only for it to be obsolete on arrival), we will aim to get 80% solutions out in the field in a fraction of that time, then continuously improve them. This is the model of software and tech innovation, and it needs to be ours as well. For example, we may launch a new drone program with off-the-shelf components to deploy within 1–2 years, then iterate newer versions every year thereafter. In practical terms, we will streamline requirements (avoiding the deadly creep of adding too many features that delay a project) and use agile development methods. Prototyping and experimentation will count as success, even if some prototypes don’t make it – what matters is learning quickly. We will also exploit the global defense market: if a foreign partner has a system that meets our needs now, we’ll buy or license it rather than reinvent the wheel. The recent creation of a £800 million UK “ARPA” (Advanced Research Projects Agency) is a step in the right direction​ (wired.com); we will leverage it to fund breakthrough ideas and accept that some projects will fail on the way to big successes. The Net Zero we seek is not carbon (in this context) but zero time wasted in fielding critical tech.

• Overhaul Acquisition Processes: We will implement the recommendations of numerous studies and committees to fix MoD procurement. This includes cutting unnecessary layers of approval, setting hard time limits for each stage of a program, and holding project managers accountable for agile delivery. For instance, if a project hasn’t shown usable progress in, say, 2 years, it will trigger an automatic review to reformulate or cancel it. We’ll also reform defence contracting to incentivize speed: contracts will include bonuses for early delivery or innovation, and penalties for excessive delay. Smaller lots and spiral development will be favored over mega-projects that aim to do it all in one go. Additionally, we will seek to “buy before we build” – meaning evaluate commercial technologies first. The old approach saw MoD trying to custom-build even basic tech, whereas now if a British company or ally already has a solution (e.g. a proven AI analytics platform or a drone design), we will adopt or co-develop it, shaving years off schedules. Another aspect is improving how we test and evaluate new tech – setting up sandbox environments (including digital twins and war-gaming with virtual prototypes) so that we can rapidly validate concepts and move them to deployment.

• Allied Collaboration and Open Markets: The UK will double down on collaborative programs with close allies to share costs and speed up innovation. We’ll actively participate in joint projects like the Global Combat Air Programme (GCAP) for next-gen fighters and look to co-develop autonomous systems with the US, Australia, and others. When appropriate, we’ll adopt systems that allies have pioneered – for example, if the US fields thousands of “attritable” drones under their new initiative​ (hoover.org), the UK might piggyback on that production run to equip our forces quickly. We’ll also open our procurement more to allied vendors, ensuring we get the best tech globally while also giving UK firms opportunities in allied projects. Small NATO members (like the Baltics and Nordics) have been nimble by buying proven systems off the shelf – we will learn from them. For instance, Estonia rapidly acquired loitering munitions and anti-tank weapons to boost deterrence instead of launching drawn-out domestic programs. Japan, in its reform, is similarly reaching out for international partnerships to bolster its defense industry and capability base​ (armscontrol.org; rusi.org). The UK will make its defense market more competitive and outcome-focused, ending the bias toward a few big suppliers who often underperform. This competitive pressure, together with allied cooperation, will spur better results faster.

• Military-Civil Fusion (Democratic Style): While we reject authoritarian China’s methods, we acknowledge the benefit of closer integration between military needs and civilian innovation. Our approach is to create a virtuous cycle between the armed forces and Britain’s world-class universities and tech companies. We will sponsor research in AI, quantum, materials science etc., with defense applications in mind, and also spin-off military-developed tech to civilian use (strengthening the economy). A key step is talent exchange: we’ll encourage secondments of tech experts into MoD and military officers into tech firms or research labs, so cross-pollination happens. The faster pace of private sector tech development must be infused into defense. Notably, Ukraine’s war effort was greatly aided by civilian technologists and apps – e.g. a civilian smartphone app was used to crowdsource targeting intel, which AI helped verify for the military​ (scsp.ai). The UK can similarly leverage civilian ingenuity for national defense (in crisis or peace) by having the right collaborations in place ahead of time. This blur of civil-military lines (in a positive sense) means when a need arises – say rapid production of drones or setting up a secure national network – the expertise and capacity of the entire nation can be brought to bear, not just the traditional defense industry.

In sum, this Industrial and Procurement Revolution frees us from the “optimized for risk-aversion” acquisition system that has held us back​ (ndupress.ndu.edu). We will be faster, more agile, and more connected with the civilian tech base. The result will be that our forces get the kit they need when they need it. Britain’s inventors and entrepreneurs will find new opportunities to contribute to national security, and the taxpayer will get better value as we avoid the costly fiascos of the past. As SCSP experts noted, we have all the resources, talent, and legal freedom needed to transform – it’s about creating the right incentives and breaking the inertia​ (ndupress.ndu.edu). This Review commits to doing exactly that, so the pipeline from concept to combat is lean and lightning fast.

Implementation and Execution Discipline

A bold plan means nothing without flawless execution. The Review thus concludes with a concrete roadmap and mechanisms to ensure these changes are realized with urgency and rigor. This is a call to action, not a theoretical paper. Key aspects of our implementation strategy include:

• Aggressive Timeline and Milestones: We will not allow this review to sit on a shelf. An implementation task force will oversee an aggressive timeline. Within 6 months, detailed action plans for each major decision (force cuts, new investments, procurement changes) will be approved by ministers. Within 1 year, we expect to see initial operational changes – e.g. at least one brigade equivalent restructured into the new model, one or two autonomous systems units formed, and legacy systems retirements initiated. By 2025’s end, the MoD should have reprogrammed funding in line with the review priorities (moving money from the cut programs to the new ones). By 2030, we aim for the bulk of the force structure changes to be complete: at least 50% of army combat units reorganized and equipped with new-generation systems, initial operational capability of key AI decision-support in headquarters, and several autonomous weapon squadrons active. We will publish annual progress reports to Parliament to ensure transparency. This wartime-like schedule is necessary – as one US defense leader said, we are in a decisive decade where delays could be perilous​ (scsp.ai). If the Pentagon is seeking to field thousands of drones in 1–2 years​ (hoover.org), we too must think in months and very few years, not decades.

• Leadership Accountability: Execution will be led from the top. The Defence Secretary will personally chair quarterly “Transformation Council” meetings with Service Chiefs and relevant industry heads to track progress. We will assign named senior champions for each major initiative (for example, a three-star officer as the AI Integration champion, another for Autonomous Systems deployment, etc.). Their task is to cut through red tape and deliver results – their performance will be evaluated on hitting the milestones. If obstacles arise (budget overruns, technology hiccups), they are empowered to find solutions or request rapid decisions from the Council. We will instill a culture where not delivering on reform is not an option. The era of military and civil service leaders being rewarded for maintaining the status quo is over; now the incentives align with driving change​ (ndupress.ndu.edu). Should there be resistance within the bureaucracy, the mandate from this highest level is clear: get on board or get out of the way. We will also engage external auditors (like the National Audit Office and independent experts) to validate that progress is real, not just on paper.

• Wargaming and Iteration: To ensure the changes are on the right track, we will continuously test and adjust our plans through wargames, simulations, and exercises. Almost immediately, we will run a series of “future force” war games that pit the envisioned new force (with drones, AI, etc.) against high-end adversary scenarios (e.g. a Baltic defense scenario against Russia, a South China Sea scenario with allies against China). These games will highlight strengths to double down on and identify any gaps or concepts that aren’t working. We’ll iterate the force design accordingly – this is the beauty of starting changes now; we have time to adjust before a real war happens. The lessons from Ukraine’s current war and other conflicts will be continually incorporated – for example, if new electronic warfare tactics emerge, we’ll adapt our drone comms on the fly. This adaptive approach guards against the plan becoming stale or misaligned; we expect the plan itself to evolve (within the Review’s intent) as we implement, guided by empirical evidence of what works.

• Resources and Budget Protection: Implementation requires funding. The government is committed to shielding and even increasing the defense budget to resource this review’s ambitions. We are aiming for defence spending nearer to 3% of GDP, following the example of Baltic states that realized higher investment is required in the new threat era​ (osw.waw.pl). However, even within current budgets, the big reallocation we outlined (cuts to old programs) frees up billions for reinvestment. We will make sure those freed funds do not slip away to other departments or get lost – they are earmarked for defence modernization. A dedicated “Future Capabilities Fund” will hold savings and channel them to priority projects (with Treasury oversight to ensure it’s spent wisely). Essentially, we’re self-funding much of the reform by cutting waste. If additional funds are needed for key leap-ahead technologies (like hypersonics or a sovereign AI capability), we will make the case to Parliament forcefully – noting that our very security and relevance are at stake. The opportunity costs of inaction (or slow action) far outweigh the costs of this transformation.

• Cultural Change and Training: Finally, we recognize that new hardware and org charts won’t succeed without a matching mindset change in our people. We are launching a concerted effort to instill a culture of innovation and adaptability in the MoD and forces. This means educating personnel about the nature of future war – helping them understand why we must change and giving them ownership of the new ideas. We’ll expand technology training for all ranks (so a young officer or NCO is comfortable using AI tools or employing drones creatively). Career paths will be modified to reward those who develop tech skills or drive innovation in their units. We’ll also address institutional inertia by rotating fresh talent into key roles – including bringing in external experts (e.g. data scientists on short service commissions) to challenge groupthink. As Kenneth Payne observed, bureaucratic inertia is a big hurdle and current organizations aren’t well-suited to the AI-driven world​ (kennethpayne.uk). We will tackle this head-on, making adaptiveness a core value. One ethos we’re promoting is “Don’t Blink” – a mantra coined by UK analysts imploring defence leaders to never look away from the rapid progress of AI and tech lest we fall behind​ (Bassett-Cross & Dear, 2024). In practice, this means our military education will include futurist scenarios, our exercises will feature new tech, and our promotions will favor those who champion improvement.

In implementing this ambitious agenda, we take inspiration from historical successes – such as the rapid mobilization and innovation in WWII, or the way some smaller nations today have leapt ahead by necessity. We are essentially forcing a peacetime revolution in defense. As Brose noted, militaries often only change under the shock of war, but the wise ones change because they see the threat looming​ (ndupress.ndu.edu). The“something worse than change” for us is clearly the prospect of the UK losing relevance or losing a conflict​ (ndupress.ndu.edu). We refuse to let that happen. Thus, execution will be relentless and unsparing. If an aspect of the plan falters, we will acknowledge it and fix it – not cover it up. If certain legacy stakeholders complain, we will listen but not be deterred from the mission. The end state we drive toward is a British military by 2030 that any potential foe knows is modern, deadly, and united– a force that can project power alongside our allies or defend the homeland against any threat, with no ifs or buts.

Conclusion: This Strategic Defence Review 2025 has pulled no punches. It spells out a transformation that is aggressive in pace and scope – because the world demands nothing less. We have identified where Britain must pivot: onto the high ground of AI and autonomy, into the new realm of kill-chain warfare, and away from the comforts of the past. We have confronted the hard trade-offs: cutting cherished programs and old force elements so that we can build the capabilities truly needed for future battles. And we have laid out how to get there, fast and sure. In doing so, we are guided by examples from forward-leaning allies and by sobering lessons (both of our own missteps and of wars unfolding in real time).

The message to all of Defence is clear: innovate and adapt, or become irrelevant. With this review, Britain chooses to adapt – aggressively, continuously, and with vision. We choose to be a leader in the revolution in military affairs, not a laggard. Our aim is that by the late 2020s, the UK armed forces will unambiguously be ready for the threats of tomorrow – a deterrent credible enough to keep the peace, and if needed, a force agile and potent enough to win swiftly. There is no room for half measures or delays. As this review has stressed, the time to act is now; the race for military relevance is on, and Britain is determined to win it​ (Bassett-Cross & Dear, 2024; politico.eu).

Britain’s soldiers, sailors, airmen, marines, and cyber warriors deserve nothing less than the best tools and a winning plan. This Strategic Defence Review provides exactly that – a ruthlessly focused blueprint to ensure that, come what may, the UK will dominate in the future battlespace.

---

So What?

Alright, tagging myself back in here. What should we make of this?

First, before you decide what you think, remember the standard you are comparing this to is twofold:

1. How far does this fall short of a perfectly written review? and

2. How much better is it than - or if your prefer how does it compare to - previous Defence Command Papers (2021, 2023).

It clearly falls well short of ‘perfection’, whatever that might mean to you. My contention would be that it is, in almost every way, better than the two most recent Defence Review papers.

I recognise it over-indexes on my own papers, but then I asked it to. That said, it offers recommendations I disagree with - this is not just the regurgitation of other’s works, or at least, no more so than something a human would write.

To take a few examples of things I don’t agree with, if it were me, I would be committing UK Defence to phasing out all crewed aircraft before 2030, to big investments in AI, and the direct pursuit of AGI, at all levels of Command and Control. I’m not convinced it makes sense to have separate Services in an age of AGI - the subject maybe for a future post. There’s nothing here on the need to reshape the MOD Operating Model. And it has dropped recommendations in the first two previous versions it produced that I thought were excellent - for example:

Accountability and Affordability: The British public and Parliament rightly demand that any extra defence spending is used wisely. To ensure accountability, the Review recommends:

• Annual progress reports to Parliament on the implementation of SDR recommendations and how budget increases are translating into outputs (e.g., number of new units formed, equipment delivered, readiness metrics).

• Strengthening the role of the National Audit Office (NAO) and parliamentary committees in scrutinizing major defence projects, to catch issues early.

• Possibly establishing an independent “Cost Realism” panel with external experts who review MoD procurement plans to ensure costings are accurate and efficiencies are on track – adding credibility to budget requests.

• Embracing greater transparency where possible on defence plans, to build public trust that increased investment is necessary and effective.

In short, it is not the Review I would have written. But the real point here is that this could be written by ChatGPT in 6-minutes. When the Review was announced in July 2024 this would have been impossible. It is worth recalling that this (too) is the worst AI you’ll ever use. By the time the Review is published it will be more capable. By the time we begin to implement the Review’s recommendations it will be better still.

Two weeks ago, researchers from around the world released a collaboration called ‘Humanity’s Last Exam’***. At the time the best AI model scored 9% in the test. Two week’s later, DeepResearch scores 26.6% (a 183% improvement in two weeks!).

Image via @kimmonismus on X.

This week, Metaculus’ forecast for the arrival of weak AGI moved from 2027 to 2026.

The conclusion should be that the MOD must address the topic of AGI in this Review. It would be a dereliction of duty to ignore it. And maybe this is Humanity’s Last Defence Review too.

---

Addendum: a further thought. You can read in my prompt that I fed Deep Research the Review’s Terms of Reference. I then added further prompts and guidance. So too in real Reviews. The reality of policy making is there will be a lot that isn’t said in the public TORs that is vital. Behind closed doors new constraints are added explicitly or implicitly but never recorded. Self-censorship takes place, where what is written is written to please seniors and stakeholders - written for what might get approved, sometimes at the expense of what needs to be said and done.

In 2019 I wrote ‘AI and Decision-Making’ for the RUSI Journal, describing how AI would drive rigour into decision-making. I argued it would force humans to make the implicit explicit, the logic, evidence and data on which decisions were made available for inspection. Which is true, and evident in the AI written Review, via the prompt, the reasoning, and the footnotes/sources. But it is more than this - it also shows how the AI-written Review was framed, what Deep Research was told it could and couldn’t say. What tone it should take. What it should seek to achieve. There’s an honesty to this, that we don’t have in human decision-making systems. I’m not sure what it means for future AI-led reviews. There would still be the Freedom of Information Act problem - if you were to make everything public including the Review’s ‘prompts’ as it were, all of its Terms of Reference, you’d end up encouraging more self-censorship from the humans involved. But given this would be across the board for all policy, maybe it will be part of much wider change in how we manage and consider information.

---

*you can find my conversation with ChatGPT’s DeepResearch, that led to this output, here.

**for more on AGI and the Review, you might be interested in my earlier piece on this ‘In Athena’s Arms’.

*** What is Humanities Last Exam? The paper describes how “High accuracy on HLE would demonstrate expert-level performance on closed-ended, verifiable questions and cutting-edge scientific knowledge, but it would not alone suggest autonomous research capabilities or “artificial general intelligence.” HLE tests structured academic problems rather than open-ended research or creative problem-solving abilities, making it a focused measure of technical knowledge and reasoning. HLE may be the last academic exam we need to give to models, but it is far from the last benchmark for AI.”

A Robot Writing a Report on UK Defence