Hackers use AI agents and get in through a vendor's backup app

Microsoft says a group used AI agents to attack hundreds of companies through ConnectWise's backup app, with no user login. What to review at your company.

On Wednesday (the 21st), Microsoft published a report from its threat intelligence team attributing a series of automated attacks with AI agents against hundreds of organizations since September to a group called Storm-0940, which it classifies as aligned with the Chinese state. The group used ARTEX, an open-source tool that puts several AI agents to work on an entire penetration test, from reconnaissance to data exfiltration, with little human intervention. According to Microsoft, the same group is behind the breach of ConnectWise Cloud Backup, a cloud backup service for Microsoft 365 and Google Workspace data. The Record confirmed the attribution, reported that ConnectWise shut down services after the alerts, and noted that the Chinese government denied any knowledge of the incidents.

How the attack worked

The attackers took control of ConnectWise’s application identities. Since those applications already had permission inside customers’ Microsoft 365 environments, all they had to do was use them to get in, without any employee typing a password or logging in.

The numbers show machine scale: access attempts against more than 2,250 customer environments, with up to 2,900 attempts per minute. Microsoft says most of them failed, but in several cases less than three hours passed between the first target and administrator access.

What this changes for your company

1. Take inventory of the keys you’ve handed out. Backup, e-signature, CRM, email marketing tools: every vendor system connected to your email and files is a key. Ask IT for a list of what each one accesses and who authorized it.

2. Least privilege and monitoring for systems too, not just people. Two-step verification protects employee logins, but it doesn’t come into play when the one accessing is an app that’s already authorized. Cut whatever is excessive and keep an eye on permission changes made by apps, such as the creation of a new administrator.

3. Prepare your response for AI speed. If the attack takes hours, the reaction can’t take days. Decide who shuts down a suspicious integration, and require in the contract that the vendor notify you quickly when it’s breached.

In 29 years of programming, I’ve learned that the most dangerous door is usually the one we ourselves left open for a trusted partner.


Originally published on IWA7 Insights, IWA7 Digital’s blog, which is now part of Escovando Bytes.

Keep readingSee all →