AI agents breach thousands of Microsoft tenants in fast-moving hack

The tension in this story lies in speed. Traditional hackers take days or weeks to move through a target’s systems. This time, a group of agents powered by artificial intelligence did it in minutes. They moved so quickly that defenders barely had time to notice they were there, leaving a trail of unauthorized access across thousands of corporate networks.

Microsoft detailed the scale of the operation in a recent security blog post. The attackers attempted sign-ins at more than 2,250 different Microsoft 365 tenants. They did this using stolen credentials from ConnectWise, a multi-tenant application. In some cases, the attackers reached privileged access levels less than three hours after they first started targeting a specific organization. The pace was relentless, with the automated system working around the clock with very little human input.

The numbers behind the attack are staggering. A separate wave of password guessing targeted 327 accounts across eight different organizations. This generated more than 151,000 failed login attempts, yet the attackers reported no successful logins from that specific method. At its peak on October 11, the system made 2,900 failed attempts every minute. For that single day, the total reached over 111,000 failures. Nearly 98 percent of these attempts were locked out by security measures, showing the system’s brute force approach.

The actor behind this is known as Storm-0940, a group aligned with the Chinese state that has been active since 2021. Microsoft describes the group as having no known industry aliases, which is typical for a developing threat actor. The industries targeted this time include aerospace and defense, semiconductor manufacturing, government bodies, and think tanks. Fourteen industries in total were on the list. It is important to note that this did not happen because Microsoft’s own systems were broken. The entry point was the ConnectWise application.

The tool they used was ARTEX, an open-source framework for autonomous penetration testing. It relies on large language models to drive the agents. These agents can scan networks, guess passwords, and abuse OAuth applications at a pace that leaves humans behind. In one specific tenant, Huntress, a security firm, watched as an app added itself to the top administrator role on October 15. This action triggered alerts for self-assigned Global Administrator roles. The attackers also planted new secrets set to expire in 2028, a move designed to maintain long-term access.

By October 16, Huntress had confirmed unauthorized sign-ins for 162 tenants. The next day, the count jumped to nearly 1,000 Entra tenants with activity from the attacking IP address, with more than 300 of those sign-ins succeeding. The pattern of the attacks, moving through tenants in sorted order just seconds apart, suggested a script working from a pre-made target list. This automated sweep allowed the group to cover massive ground in a very short time window.

The use of ARTEX is a shift from traditional manual hacking. The framework can use any AI model, and threat actors in multiple regions are adopting it. Recent reports tied similar technology to breaches at South Korean financial institutions. The developer of ARTEX, known by the handle Autumn-27, announced on GitHub that the project will no longer be updated and will become closed source. This development could change how other attackers use the tool going forward.

ConnectWise has suspended several of its services as a precaution, including Cloud Backup and SaaS Security. They are working on a phased restoration that is expected to take several days. There are currently no indicators that partner backup data was compromised. Microsoft advises customers to keep affected app connections disabled until official instructions are issued. They also recommend reviewing activity logs from this month to see if any unexpected sign-ins occurred. The key takeaway is that AI is making supply chain attacks faster and harder to stop.

This incident highlights a growing risk in the digital supply chain. When a single third-party tool is compromised, the damage can ripple out to thousands of customers. The use of autonomous AI agents changes the timeline for defenders. They can no longer rely on slow, methodical human attackers. The clock is ticking faster, and the window to respond is shrinking with every new tool that hits the open-source market.