This is the online edition of The Wiretap newsletter, your weekly digest of cybersecurity, internet privacy and surveillance news. To get it in your inbox, subscribe here.

The Cybersecurity and Infrastructure Security Agency (CISA) is finally using Anthropic’s Mythos AI model to search for vulnerabilities that could have a serious impact on government agencies and private businesses, per a report in Reuters.

Forbes previously reported that CISA didn’t have access to either Anthropic or OpenAI models that had proven adept at finding bugs and exploiting them with no human required. Anthropic has limited Mythos access to select partners, but CISA was not one of the first cohort of nearly 50 organizations.

Now, though, as confirmed to Forbes by an anonymous CISA employee, some staff on the Attack Surface Evaluation team have been using the tool to scan government code stored on Github. Reuters, citing three people familiar with the matter,........

© Forbes