Mandiant reports that attackers took over a live AI coding-assistant session inside an unnamed SaaS company and used it to plant the Shai-Hulud worm across roughly 100 internal repositories. The intrusion began when the compromised assistant suggested a poisoned software package that a developer accepted, giving the attacker a way to spread further. From there, the worm went on to steal repository secrets and source code, highlighting the danger of trusting AI coding suggestions without verification.
// source: thehackernews.com ↗