Anthropic delays AI model it deems too powerful for public use

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.

Tech firms are known for hyping their own products, but Anthropic’s decision to delay the release of its new AI model has turned heads. Anthropic said that Claude Mythos Preview is able to identify and exploit weaknesses in critical software programs with simple prompts. The company claimed that Claude had found bugs and vulnerabilities in nearly every major operating system and browser, including those designed to be difficult to hack. Instead of putting the software in the public domain, Anthropic will make it available to a select group of companies which will use it to fix security issues in their products. Although the popularity of Claude is down to its adeptness at coding, its other skills mean it could be a dangerous tool in the hands of adversaries. Hackers will be doing their best to get hold of it.