New claims about Anthropic’s AI model set alarm bells ringing

This article appeared as part of the Daily Sensemaker newsletter – one story a day to make sense of the world. To receive it in your inbox, featuring content exclusive to the newsletter, sign up for free here.

Allegations that a handful of people on a private internet forum have gained early access to Anthropic’s Claude Mythos model have set alarm bells ringing. The tech company has confirmed that it is now investigating the claims. Although Claude’s large language models have been widely disseminated, the newest version has not been publicly released. Anthropic claims Mythos can easily hack and discover cyber-security weaknesses, and said earlier this month that it had already found thousands of vulnerabilities “in every major operating system and web browser”. Britain’s AI minister says that Mythos could present a danger to UK businesses if the tech falls into the wrong hands. The potential dangers could go well beyond this.