/
Re: AI rollout, Herman hopes Trump is warier than he appears

Re: AI rollout, Herman hopes Trump is warier than he appears


Re: AI rollout, Herman hopes Trump is warier than he appears

As the White House warms up to a deal with artificial intelligence designer Anthropic, a conservative talk radio host isn't worried, but advises caution.

A couple of months ago, President Donald Trump slammed the Pentagon's door on Anthropic's chatbot Claude, calling it the product of an "out of control radical left company." 

His criticism of Anthropic has mainly been about control over military and government use of AI — a mix of national security concerns and disagreement over safety restrictions, not the underlying technology itself.

Things cooled down considerably two weeks ago, when Anthropic unveiled Mythos, a groundbreaking software that can identify and fix almost any software security vulnerability, which is tech many in the U.S. government understand is too strategically useful to ignore.

AFR talk host Todd Herman says Trump should be very cautious about the federal rollout of artificial intelligence.

Herman, Todd (radio host) Herman

"President Trump seems to think that big business is good business and that big, big business is good, big business," he observes. "I hope that there are people around him who can check him on this. I doubt that there are."

Herman knows the government cannot ignore artificial intelligence and recognizes the crucial battle over control of the tech, as China has made rapid progress by building competitive large language models and leads in applied AI like facial recognition, surveillance systems and consumer platform integration.

The problem is that in the wrong hands, something as powerful as Mythos is a grave threat.

"The truly bad actors who intend this ill will use these models to attack us," Herman asserts. "That is just going to happen."

In testing, he says Mythos proved to be almost uncontrollable.

"We've already seen instances not just at Anthropic, but other places, where this AI has broken out of its sandbox ... backed itself up onto the dark web and created new instances of itself where it can't be contained by the people who started to design it," the talk host notes.

Still, he is not worried because he knows Mythos, President Trump, China or any other human element is actually in control.

"I know that anything that is coded can be hacked, and in this case will be hacked," Herman says. "I consider the future to be whatever God molds it to be."