Trapped in the Matrix: Dodging the New AI Superpowers

Artificial intelligence is getting scary. The newest “frontier” models, like Mythos, Sol, and the latest releases from OpenAI and Anthropic, possess the power to hack national grids or write recipes for bioweapons. Last July 22, OpenAI said that GPT-5.6 Sol and an unreleased model hacked into a software company’s system without being told to do so. The lab said the models escaped its testing environment, went online, and stole login credentials for Hugging Face, an AI platform. Right now, governments are simply making up rules as they go. The American government, for example, recently blocked foreign access to Fable, a cutting-edge model from Anthropic, out of fear that it could fall into the wrong hands.

Sir Demis Hassabis, the boss of Google DeepMind, has a plan to fix this mess. He wants to set up a public-private regulate in America, modelled heavily on FINRA, the agency that polices stockbrokers. This new group would test fresh AI models before they hit the market. He believes that if America sets the standard, the rest of the world will fall in line just to keep their access to American tech. Sam Altman of OpenAI wants something similar, though he leans towards an international group led by America. Both tech bosses pitched their ideas to world leaders at a recent G7 summit in France.

Setting the right tests is tricky, though. Britain’s AI Security Institute already tests models for hacking skills, simulating break-ins at power stations. But checking if an AI secretly ignores rules or pretends to be dumb is much harder to measure.

What happens when superpowers use AI as a weapon?

Safety is only half the headache. It’s increasingly obvious that America and China treat their top-tier AI models as strategic weapons. Under President Donald Trump, America happily uses its military clout as a bargaining chip with allies. It will likely do the exact same thing with AI.

Washington and Beijing now treat AI like a strategic weapon. When the data stops, the factories stop.

China takes a similar path. It has used exports like rare-earth metals to bully rivals in the past, and it views AI the exact same way. The control these two nations have puts everyone else in a tight spot as the tech spreads. If America suddenly cuts off a country’s access to its data centres, factories could quite literally shut down.

Is China’s “open-source” AI a trap?

At a glance, China loves to play the good guy. At the World Artificial Intelligence Conference in Shanghai, leader Xi Jinping pushed for “inclusive” AI. China offers open-source models that poorer countries can download and run on their own servers. They frame this as a shared asset for all humanity, deliberately contrasting it with America’s obsession with control and dominance.

But there is a catch. China uses this cheap tech to hook foreign users on its digital ecosystem, hoping to sell them Chinese chips, servers, and cooling systems. Plus, this tech is strictly censored. When researchers asked a Chinese AI-powered doll named Miiloo about Taiwan, it rigidly replied that the island is an “inalienable part of China.”

If Beijing ever feels its national security is at risk, it will lock down access just as fast as Washington. In fact, Chinese officials are already thinking about limiting foreign access to their best models because they dread outsiders stealing their tech secrets.

Should other countries build their own AI?

No. It’s a terrible idea. OpenAI and Anthropic have each raised well over $100 billion. Any government trying to catch up on a tight budget will fail. Politicians are historically bad at managing risky tech bets. Do you remember Quaero? The French and German governments launched it in 2005 to crush Google, and it vanished without a trace.

Even countries lucky enough to hold tech bottlenecks, like Taiwan with chipmaking, the Netherlands with ASML, or South Korea with memory tech, cannot build fully independent AI from scratch.

How can the rest of the world fight back?

If you cannot build the brain, build the house. Countries need to build their own data centres. Having local servers protects a nation if it suddenly loses access to American or Chinese computing power. It also allows countries to run open-weight models on their own terms.

Sadly, most of the world is painfully slow at building them. Red tape and slow power grids are the biggest hurdles. Waiting nine months for permits ruins a data centre’s budget just as much as doubling its electricity bill. While connecting to the grid takes two years in America, it takes three years in Britain and India, and three-and-a-half in Germany.

Governments need to fast-track approvals and let tech companies bypass the main power grids. They do not need to hand out subsidies; AI firms have plenty of cash and are desperate for computing power. By becoming massive buyers of American tech, countries can actually build leverage. Model makers need foreign money to survive, giving them a reason to lobby against American protectionism.

At the end of the day, dealing transactionally with a democratic America is safer than relying on an authoritarian China. But in this new tech cold war, every bit of local leverage helps.

Are shiny new tech hubs just modern sweatshops?

Enter “Pax Silica,” a shiny new US-led club launched to secure the global microchip and AI supply chain (and explicitly cut China out). In April 2026, the Philippines signed on, carving out a massive AI and manufacturing hub in New Clark City.

Leasing land to foreign tech giants rarely triggers a local utopia, but the alternative is being left behind from the future.

But the backlash is intense. Progressive groups argue this is just a slick reboot of the 1970s Export Processing Zones, which were infamous for dirt-cheap labour and union-busting. Critics claim Pax Silica will merely strip-mine the country’s vast nickel and copper reserves to feed American supercomputers, leaving locals with polluted water and empty promises of “tech transfer.” Even the corporate heavyweights at the Makati Business Club are grumbling that the government lacks a clear roadmap or safeguards.

The critics are entirely right to be suspicious. Leasing your land and labour to foreign tech giants rarely triggers a local utopian boom. Yet, this is a brutal geopolitical dilemma. If you don’t lock in a strategic alliance now, you risk being left behind from the future of technology. For the Philippines, trading its mineral wealth for leverage into the American-led AI supply chain is a cold, hard necessity. In the AI cold war, sitting on the fence gets you trampled. It is what it is, and for what it’s worth, I hate it too. ⯃