US speeds up AI for national security, asks top AI models to pass gov safety tests first

The US government said Friday it will speed up AI development for national security, while making clear the tech should not be used for illegal spying. Earlier this week, the US asked top AI companies to voluntarily let the government test their most powerful models before releasing them to the public.

In a national security memo, President Trump wrote that America "can and will" responsibly speed up AI use in intelligence and combat, based on American values. He ordered the Defense Secretary to update rules on autonomous weapons within 90 days, to make sure AI systems are used carefully and respect the chain of command. The memo also says national security agencies must not use AI to "censor free speech" or carry out unauthorized spying.

White House tech chief Michael Kratsios said the memo pushes for AI systems from multiple suppliers to avoid a single point of failure, and updates Pentagon guidelines on autonomous weapons. This comes as AI lab Anthropic clashes with the Pentagon. In March, the Pentagon labeled Anthropic a supply chain risk after it refused to lift bans on using its Claude model for autonomous weapons and mass surveillance inside the US.