The debate over artificial intelligence is entering a more consequential phase. What began as a contest over computing power, model performance and venture capital is increasingly becoming a struggle over regulation, national security and technological sovereignty.

Anthropic’s warnings about the risks posed by increasingly capable AI systems have intensified that debate, but a new analysis raises an important caution.

Genuine concerns about AI safety should not become a justification for entrenching America’s leading laboratories or deepening technological confrontation with China.

The distinction matters because AI safety is both a technical problem and a geopolitical one. If governments respond to frontier-model risks by concentrating access to advanced chips, computing infrastructure, talent and research within a small group of US companies.

They could reduce certain risks while simultaneously creating a more concentrated technological order. Such concentration may give established laboratories greater influence over the rules governing an industry that is becoming critical to economies and governments worldwide.

Anthropic has been among the companies publicly warning that frontier AI could eventually generate risks that existing institutions are poorly equipped to manage. Those warnings deserve serious consideration.

Particularly as AI systems become capable of operating with greater autonomy, using tools and interacting with complex digital environments. But the policy response must distinguish between legitimate safety measures and policies that primarily protect incumbents from competition.

This is where Sebastian Mallaby’s discussion of AI regulation becomes significant. The author of The Infinity Machine has examined how technological revolutions can reshape economic power, financial markets and institutions.

The emerging AI transition presents a similar challenge: regulation must become strong enough to manage systemic risks without becoming so restrictive that it suppresses experimentation, competition and technological diffusion.

For Europe, the question is particularly difficult. The continent does not possess the same concentration of frontier AI laboratories or hyperscale computing capacity as the United States.

Nor does it occupy China’s position as a major state-backed technological competitor. Yet Europe holds other cards: regulatory influence, industrial expertise, large consumer markets, research institutions and the ability to establish standards that companies must follow if they want access to European customers.

The European Union’s AI regulatory framework demonstrates how that influence can operate. Rather than attempting to dominate the frontier-model race purely through scale, Europe can shape the conditions under which AI is deployed.

Requirements surrounding transparency, risk management, accountability and data governance could become commercially important if global companies design products around European standards.

But regulation alone will not secure Europe’s technological position. Excessive compliance costs could discourage startups and push investment elsewhere, while insufficient investment in computing infrastructure could leave European researchers dependent on foreign platforms.

Europe therefore faces a dual challenge: regulate powerful AI systems while simultaneously building enough domestic capability to remain strategically relevant.

The same tension applies to the US-China relationship. Treating AI exclusively as a national-security competition risks transforming safety policy into another mechanism of technological decoupling.

Yet ignoring geopolitical competition would also be unrealistic. Advanced AI has implications for economic productivity, cybersecurity, military capabilities and scientific research. The central challenge is therefore institutional rather than merely technological.

Governments need mechanisms capable of monitoring frontier systems, responding to demonstrable risks and coordinating internationally without allowing safety concerns to become a blanket argument for market concentration.

AI regulation will be judged not only by how effectively it prevents dangerous outcomes, but also by whether it preserves competition, innovation and international cooperation. The future of AI should not be determined solely by whichever laboratories build the most powerful models first.

It will also depend on whether governments can construct rules that make technological power accountable without turning legitimate safety concerns into instruments of geopolitical escalation.

How Geopolitics, Bond Markets, AI and Infrastructure Are Reshaping Business

The global economy is not simply experiencing another cycle of uncertainty. It is undergoing a repricing of what businesses consider valuable.

For decades, globalization rewarded companies that could produce more cheaply, borrow more cheaply and operate with remarkably lean supply chains.

That model is being challenged by war, volatile capital markets, energy insecurity and the rapid construction of artificial-intelligence infrastructure. The emerging economic argument is distinctive: resilience is becoming an economic asset, not merely a defensive expense.

Geopolitical conflict has exposed the hidden cost of efficiency. A supply chain designed around the cheapest supplier can become extraordinarily expensive when shipping lanes are disrupted, sanctions are imposed or critical components become unavailable.

Companies are consequently accepting higher short-term costs to gain greater control over production and sourcing. Factories are being diversified, suppliers are being duplicated and strategic inventories are being reconsidered.

This represents a fundamental change in corporate economics. Redundancy was once treated largely as waste. Increasingly, it is being treated as insurance. A second supplier, a domestic production facility or a larger inventory may reduce margins in normal conditions while protecting revenues during a crisis.

The question facing executives is therefore changing from “How cheaply can we operate?” to “What level of disruption can our business absorb?” Energy illustrates the same transformation.

Oil-price volatility and geopolitical tensions have made energy security an increasingly important component of industrial strategy.

Companies cannot easily separate production costs from global political developments when fuel, electricity and transportation are exposed to international shocks.

This is helping strengthen the economic case for alternative energy sources, efficient infrastructure and long-term power agreements. Bond markets are revealing another side of the adjustment.

Governments and corporations are operating in an environment where capital cannot be assumed to remain permanently cheap. Elevated borrowing needs, inflation uncertainty and changing expectations about monetary policy can push long-term yields higher even when economic growth remains subdued.

That creates a more demanding environment for investment: projects must generate convincing returns rather than depending on inexpensive financing to make the numbers work.

Artificial intelligence complicates the picture because it is simultaneously a technology revolution and an infrastructure boom. The enormous investment required for data centers, chips, electricity generation and networks represents a bet that future productivity will justify today’s capital expenditure.

The crucial economic question is therefore not whether AI is transformative, but whether its productivity gains will become large and widespread enough to support the valuation of the infrastructure being built around it. This distinction matters.

An economy can experience an investment boom without immediately experiencing a productivity boom. If AI substantially increases output per worker, it could support faster growth while easing some cost pressures.

If adoption remains concentrated in a limited number of highly profitable companies, the benefits may be less broadly distributed. Infrastructure itself is consequently becoming a strategic economic variable.

Electricity grids, semiconductor capacity, ports, telecommunications and digital networks are increasingly viewed not simply as background utilities but as productive assets that determine how quickly economies can respond to technological and geopolitical change.

Meanwhile, consumers remain the test. Higher financing costs and persistent price pressures can eventually weaken household demand, forcing companies to confront slower revenue growth alongside higher operating and capital costs.

The defining economic story, then, is not merely disruption. It is repricing. The world is placing a higher value on security, flexibility, reliable energy, technological capacity and strong balance sheets.

Companies that once competed primarily through efficiency are increasingly competing through their ability to remain operational when efficiency alone is no longer enough.

The next phase of globalization may therefore be less about minimizing every cost and more about determining which costs are worth paying to preserve economic control.