Two weeks ago, Manchester Mayor Andy Burnham and former Bank of Canada Governor Mark Carney stood on a stage and nodded solemnly about AI risk and defense cooperation. The headline wrote itself: two heavyweight policy minds charting a transatlantic alliance to tame machine intelligence and secure our borders.
It sounds responsible. It sounds statesmanlike. It is also a complete waste of time.
Whenever politicians gather to wring their hands over artificial intelligence safety and cross-border defense interoperability, they are treating a symptom while ignoring the disease. They want you to believe that the primary danger of advanced automation is rogue algorithms taking over military command chains or algorithmic bias in hiring. That is the safe narrative. It gives regulators something to regulate, consultants something to bill for, and politicians a reason to look grave in front of cameras.
I have spent the last decade watching institutional boards burn hundreds of millions of dollars trying to future-proof operations against hypothetical machine uprisings while their core infrastructure rotted from bureaucratic cowardice. The real crisis is not that our models are too smart. The crisis is that our institutions are too slow, and our leaders are terrified of the economic disruption required to actually compete.
Let us dismantle the polite fiction of the Burnham-Carney consensus.
The Defense Procurement Mirage
Start with the defense cooperation angle. For decades, the Anglo-Canadian military-industrial complex has relied on a cozy club of prime contractors, decades-long procurement cycles, and committee-driven requirements. When politicians talk about integrating AI into defense, they imagine a sleek command center where a British automated drone network coordinates with Canadian surveillance satellites over the Arctic.
The reality on the ground looks entirely different.
Walk into any defense ministry in London or Ottawa today, and you will find engineers trying to bolt advanced machine learning models onto legacy hardware running on software architectures older than the programmers writing the patches. You cannot solve a software problem with a hardware budget.
Carney and Burnham talk about sharing intelligence and aligning regulatory frameworks for defense tech. But while western governments spend four years debating the ethics of autonomous targeting systems, adversarial state actors are buying commercial off-the-shelf hardware, open-source weights, and deploying decentralized drone swarms at a fraction of the cost.
The bottleneck in modern defense is not capital, and it is not lack of cooperation treaties. It is risk aversion. Our procurement processes are legally optimized to prevent scandal rather than to win conflicts. Until we accept that high-stakes technological competition requires accepting high rates of failure, every joint defense communiqué is just expensive paper-shuffling.
The Artificial Intelligence Safety Theater
Then there is the AI risk conversation. For the past three years, the dominant discourse has been hijacked by existential panic. We are told that soon, frontier models will achieve general intelligence, outsmart their creators, and present a civil existential threat unless we establish immediate international oversight bodies.
This is regulatory capture dressed up as altruism.
When legacy economies lobby for heavy compliance frameworks, licensing regimes, and centralized review boards, they are not protecting the public. They are pulling up the drawbridge. Large technology incumbents love safety debates because compliance costs crush garage-startup competitors while leaving big tech's multi-billion-dollar clusters untouched.
Andy Burnham wants to position northern England as a tech hub, and Mark Carney understands global macroeconomic capital flows. Yet both miss the core economic reality: artificial intelligence is not a product category you regulate like pharmaceuticals. It is a general-purpose technology, akin to electricity or internal combustion.
Trying to govern AI risk by focusing on the model weights is like trying to stop car accidents by regulating how refineries process crude oil. You regulate the deployment and the liability, not the foundational math. If a model causes harm in a medical diagnosis or a financial loan denial, the institution deploying it must face catastrophic legal liability. That single policy shift would clean up the industry overnight far better than any international safety summit ever could.
The Capital Allocation Fallacy
Mark Carney made his name steering central banks through crises by demanding long-term thinking and green capital allocation. The assumption in his circles is that private markets are failing to fund the right kind of AI because short-term investors only want software that writes marketing copy or generates cat pictures.
The inverse is true. Private markets are drowning in capital looking for yield. The reason we are not seeing productive breakthroughs in areas like materials science, nuclear fusion optimization, or logistics is not a shortage of venture funding. It is a shortage of institutional permission to fail.
When a startup builds a foundational model, investors demand immediate SaaS metrics. When a legacy corporation tries to integrate machine learning into supply chains, middle management protects their headcount by burying the project in pilot purgatory. We have built an economy where financial engineering pays better than operational engineering, and no amount of transatlantic dialogue between former central bankers will fix that cultural rot.
What Real Competitiveness Looks Like
If we want to survive the next decade of geopolitical and technological friction, we need to stop pretending that committees and shared declarations are strategy.
- Abolish legacy procurement. Defense ministries must adopt a venture-capital model where they fund dozens of unproven software and hardware startups with the explicit expectation that eighty percent will fail.
- Shift from input regulation to liability enforcement. Stop worrying about what models can think and start punishing organizations brutally for what automated systems actually do.
- Decouple industrial policy from political geography. Innovation clusters do not happen because a mayor designates a district as an innovation zone; they happen when talent and capital are allowed to collide without municipal interference.
The conversation between Burnham and Carney is comfortable because it assumes the current structure of western governance is sound and merely needs fine-tuning for the twenty-first century.
It is not sound. The machinery is broken, the incentives are inverted, and the threats we face do not care about joint communiqués.
Stop looking to politicians to save us from technology. Start demanding that they get out of the way of the people actually building the future.
