Attackers have built an AI tool that watches popular web software for new patches, works out what each patch fixed, then writes an exploit and runs it without a person involved, according to Google’s Threat Intelligence Group, whose researchers say they hadn’t seen the whole process automated before. One example in the report shows the tool, which its configuration calls exploit_agent, checking access controls in an older version of Spectra, a WordPress plugin with more than a million active installs. Researchers found it on an exposed server holding stolen data and webshell scripts, with victim records pointing to French sites in media, entertainment and education, yet they haven’t tied a specific breach to the tool or named who runs it, despite Russian comments in its configuration.






