Everyone vs. Superintelligence?

“Et tu, Satya? It must be international "gang up on frontier models" week. The Trump administration still hasn't lifted its effective ban on commercial distribution of the latest model from Anthropic, America's leading AI company. Microsoft CEO Satya Nadella added his own kick in the shin, calling for cheaper, decentralized AI to undercut the startups leading (and funding) a quest for artificial superintelligence. If that wasn't enough, Five Eyes -- the spy-agency collaboration of the U.S., U.K., Canada, Australia and New Zealand -- issued a rare public warning about the dangers of fast-advancing artificial intelligence.

 

Unrelated or not came Tuesday's global tech selloff, an aggressive Chinese manifesto on AI competition and a renewed Beijing rare-earth war against Western tech manufacturers.

 

We're just under two weeks into a Trump-inaugurated new artificial-intelligence era. The most cutting-edge products are too dangerous to be released to the public. Governments will treat them like weapons that only government should have access to. When the Trump administration reacts fearfully to the rise of a company like Anthropic, incoherence and chaos are the Trump way. But any administration would react. When government says it's worried about public safety and national security, in both cases it's worried about its own authority. It's worried about the government not being the government anymore.

 

You can see the same thing happening this week in Europe. Its defense thinkers are raging quietly about the continent's strategic vulnerability after the U.S. cut off access to Anthropic's Fable 5 model.

 

But without customers, how are America's AI leaders supposed to keep funding the push toward ever more powerful artificial general intelligence? Will IPO investors still come as expected to fill the hole?

 

Don't be so sure. Four months ago Anthropic's Dario Amodei could comfortably speculate that "things would have to go pretty badly" for the industry's trillion-dollar investment not to be quickly repaid with trillion-dollar revenue. "If you're off by only a year," he added, "you destroy yourselves."

 

Well, a definition of "pretty badly" is government ripping your latest product out of the hands of millions of eager customers.

 

Not that the Trump administration's fears are a figment of an overactive imagination. Mr. Amodei himself has publicly fretted about terrorists, criminals and despots getting access to what he calls 50 million Nobel Prize-level geniuses in a data center -- i.e., the wherewithal to manufacture a mold-breaking chemical, biological or cyber weapon.

 

Other fears, such as destabilizing job loss, strike me as overblown or unduly speculative. The so-called alignment problem, or AIs developing their own agendas and deciding to eliminate humanity? I worry about that even less: Governments clearly will stamp down on anything they don't control.

 

In seizing control, though, they can also stop progress that already faces challenges on the financial front. Last year, the entrepreneur behind China's DeepSeek paid lip service to the costly quest for superintelligence. But he already was pricing his model to undercut Western models, which he was suspected partly of having pirated. Now China's agenda has become Microsoft's agenda. Ditto many startups devoted to mining the vast gap between what it costs the leading AI companies to produce their research and what customers are willing to pay for access to the models.

 

Pete Hegseth at the Pentagon may nurse an infantile grudge against Anthropic, accounting for some of the company's troubles with the administration. But wake up to reality. Being negotiated in real time is either the end of U.S. progress toward superintelligence, because investors will stop funding it, or the end of such advances being available to anyone except the U.S. government and its designated licensees.

 

Yet this is only a stopgap in addressing the larger problem. How to curb rogue actors in the long run? How to establish AI deterrence against China, Iran, Russia and North Korea? It's already clear that AI will have to be the solution to problems created by AI. Given the inherently unpredictable and spontaneous nature of AI outputs, experts also doubt that built-in safeguards (i.e., censorship) in publicly released models can be a sufficient fix. That leaves one path.

 

In past years this column has suggested how surveillance might address America's mass-shooter problem. The data only needs to be aggregated; red-flag algorithms need to be programmed in. In the future, an irreducible function of AI will certainly be to monitor how people and governments around the world are using AI to identify and interrupt antisocial projects before they come to fruition.

 

A coming surveillance state has seemed inevitable. The question has been how to contain it within democratic, legal and constitutional constraints. The time has come to get cracking on this problem. It may be the best or only way to make sure AI gains can then be widely implemented in the economy.” [1]

 

1. Main Street: Everyone vs. Superintelligence? Jenkins, Holman W; Jr.  Wall Street Journal, Eastern edition; New York, N.Y.. 24 June 2026: A13.