Criminals Are Using Self-Running AI to Attack Financial Firms, Microsoft Says
Microsoft Threat Intelligence says a financially motivated criminal group it tracks as Storm-3358 began attacking organizations in the financial sector in October 2026 using ARTEX, an offensive framework that lets AI agents carry out attack steps without a person directing each one.
In a post published Nov. 6, 2026, the company says ARTEX agents were used alongside conventional automated tools to scan systems connected to the public internet, exploit security flaws and install open-source remote-access trojans. Microsoft says tools of this kind can use stolen credentials and compromised systems at machine speed to move through a network, gain higher privileges and pull data out. It says the activity fits what it calls an ongoing trend of attackers accelerating every stage of an attack with agentic AI frameworks.
Microsoft describes Storm-3358 as a Cambodia-based intrusion cluster that mainly targets financial organizations in Southeast Asia, along with a smaller number of large financial organizations elsewhere. It says the group used three open-source remote-access programs to hold access and move inside victim networks: WISP, vshell and AdaptixC2.
The company's defensive advice targets the automation itself. Microsoft says an agent-driven scan sends many different requests in a short time and retries variations or moves to its next step at machine speed. Its guidance recommends honeypots, decoys, tarpits and what it calls "fake wins": worthless targets that spend the agent's time and compute budget, or convincing evidence that the job is already done, so the agent stops on a false success. Microsoft says those controls cut the attack traffic reaching real assets in its own testing.
Storm-3358 is a different group from the one in the ARTEX campaign reported in October, which Microsoft tracks as Storm-0940 and which hit a different set of victims.
Microsoft published sample indicators of compromise and detection guidance with the post, and said attackers routinely change the source address of malicious traffic to get past static detections.
