It’s called “Mythos,” and it’s the most disruptive artificial intelligence (AI) model yet developed.
The government claims that Anthropic’s Mythos is a “generational leap” beyond other AI models. Washington believes that Mythos is an AI capable of “not just identifying weaknesses in security systems, but exploiting them with autonomous, never-before-seen precision,” reported Axios.
That’s why Anthropic gave Mythos a “limited release” to about 40 companies so they could develop defenses against it before it was released into the wild.
On the heels of Mythos, ChatGPT 5.5 is also being tested. These advanced AI models have the government doing an about-face and seriously considering far stricter oversight of AI models before they’re sold to local, state, and national governments.
“The Trump administration is considering an executive order to establish a working group on AI featuring tech executives and government officials, including plans for a formal government review process for new AI models,” the New York Times reported.
Last year, the administration appeared to be signaling that the government would regulate AI with a light touch. “We have to grow that baby and let that baby thrive,” Donald Trump said last year, adding, “We can’t stop it with foolish rules and even stupid rules.”
Then came the war between the Pentagon and Anthropic over the company’s refusal to grant the military unfettered access to Claude, Anthropic’s top AI model. Anthropic’s co-founder, Jack Clark, feared the military would use Claude for surveillance and autonomous weapons. Eventually, the Pentagon ordered Claude to be ripped out of every military computer, and Secretary of War Pete Hegseth designated Anthropic a “supply chain risk.”
Anthropic has since entered into negotiations with the Pentagon to allow the military to use Mythos. Last week, the Pentagon announced agreements with OpenAI, Alphabet, Nvidia, SpaceX, Microsoft, Amazon, and Reflection for their AI models to be used on the agency’s classified networks for “lawful operational use.”
In March, Trump appointed 13 members to a new AI advisory panel, whose members reportedly included Meta’s Mark Zuckerberg, Oracle’s Larry Ellison, Nvidia’s Jensen Huang and Dell founder Michael Dell. Trump said in January the group would advise the president on “matters involving science, technology, education and innovation policy.” The appointments came as the Trump administration issued a legislative framework it wanted Congress to implement, creating a uniform national policy on AI that would supersede state rules.
“Government pre-release review mechanisms, as seen in U.K. proposals, typically center on structured risk assessments, adversarial testing, and controls around models with high-capability outputs. Industry practitioners confronting comparable frameworks usually need stronger red-teaming workflows, reproducible evaluation artifacts, and clear data-handling protocols to satisfy external assessors,” according to Let’s Data Science. Safety protocols that include strict data-handling requirements and the testing agency’s ability to reproduce results from the company’s tests are the absolute minimum for adequate AI vetting.
It’s a legitimate question to ask whether this executive oversight of AI models is necessary and whether it will hurt us in competition with China. Obviously, it depends on the kinds of oversight sought in Trump’s executive order. What causes me less anxiety is that I’d rather Trump be making the rules than some Democrat luddite who listens to radical left scary scenarios based on a movie script rather than cold, hard science.
Editor’s Note: Do you enjoy PJ Media’s conservative reporting that takes on the radical Left and woke media? Support our work so that we can continue to bring you the truth.
Join PJ Media VIP and use promo code FIGHT to receive 60% off your membership.
