Building powerful AI tools is the goal of companies like Anthropic. But after realizing how powerful its latest update was, Anthropic moved to keep it from the public and turn it into a cybersecurity stress test.

On today’s Big Take podcast, Bloomberg cybersecurity reporter Margi Murphy walks host Sarah Holder through the story of how Anthropic discovered the risks of its Mythos tech and what they could mean for banks, businesses and the government.

How Anthropic Learned Mythos Was Too Dangerous for the Wild
Anthropic’s Mythos Model Is Being Accessed by Unauthorized Users

We have a special Bloomberg subscription offer for podcast listeners at Bloomberg.com/podcastoffer.

Hosted by Sarah Holder; Produced by Julia Press; Reported by Margi Murphy; Edited by Jeffrey Grocott.

Fact-checking by Eleanor Harrison-Dengate; Engineering by Alex Sugiura.

Senior Producer: Naomi Shavin; Deputy Executive Producer: Julia Weaver. Executive Producer: Nicole Beemsterboer.

See omnystudio.com/listener for privacy information.

An AI So Powerful Anthropic Kept It From the Public