first the Pentagon, then the rest of the world

The United States had a maxim to win the AI ​​race: that there were no rules or limits. Politicians and regulators have turned a blind eye because what mattered was that US companies could develop the best AI models without restrictions. The problem is that some models have ended up being so good that now the US Government is beginning to fear their potential. Your easy solution: review them before anyone can use them.

Is the open bar over? Donald Trump’s team is designing a plan intended to have a formal supervision process for new artificial intelligence models. Under this structure, a group of experts and government officials would analyze and review each new model and approve its launch. The implications are enormous, because the US Government would have preliminary and exclusive access to the model before its mass launch.

Trump did not want restrictions on AI. Last summer, Donald Trump compared to AI with a “beautiful baby that has been born. We have to grow that baby and let it thrive. We can’t stop it. We can’t stop it with politics. We can’t stop it with absurd rules, not even stupid rules.”

But he’s staying alone. This total support for AI is leaving the US president in an uncomfortable position. Both Democrats and Republicans they are worried over the risks posed by AI, and a Pew Research Center survey from last year Indian that half of all of them do not welcome how AI is increasingly used on a daily basis.

Mythos as a turning point. This feeling of rejection has been growing among political groups, but fears have been reactivated especially after the launch of Claude Mythos Preview by Anthropic. The company only allowed access to the model to a small group of technology partners, claiming that it was too advanced in areas such as cybersecurity. The internal tests Of course They seemed to demonstrate their potential.

Priority access. The White House wants to avoid political repercussions from a potential cyber attack created by AI, but at the same time the administration is assessing how these capabilities can be useful to the Pentagon and US intelligence agencies. Some government officials want this review system to be put in place that allows priority access to these models even if that does not block their launch.

Good news for China. This type of plan has a big problem: it would slow down the development of new models and the innovation capacity of US companies compared to Chinese companies, which do not undergo this type of prior government review. And if they do, it is a problem that is not reflected in the advancement of the AI ​​models of Chinese companies, which are increasingly closer in terms of capacity to the best models of US companies.

There is no more “AI czar”. These plans coincide with another unique event. In March, David Sacks, the so-called “AI czar” of the White House, left his position. He has been replaced by Susie Wiles and Scott Besset. While Sacks fully supported such a “no rules” policy for AI, both Wiles and Bessent intend to have more control over the policies that apply to AI.

AI as a weapon. All events make it inevitable to compare the development of AI with the development of the atomic bomb in World War II. It was the US Government that led and controlled this technology in the past to have a definitive strategic advantage against its adversaries, and the same could happen with AI now that it is becoming a potential “cyber weapon.”

In Xataka | Only a handful of US companies have access to Claude Mythos: the ECB already fears for the savings of all of Europe