Security · November 20, 2026

Also published in Português (Brasil)

Federal Bureau of Investigation warns of artificial intelligence tool misused in cyberattacks penetrating systems in three hours

A server room with rows of black and white cabinets under blue cables
Taylor Vick / Unsplash

The Federal Bureau of Investigation has warned that hackers are exploiting an open-source artificial intelligence security testing tool called ARTEX to breach organizations worldwide. According to the agency's alert released on November 13, attackers have used the tool to gain privileged access to target systems in as little as three hours after selecting their targets.

The Federal Bureau of Investigation stated that attackers have employed ARTEX since at least August to target banks, government agencies, technology companies, and healthcare organizations. The tool was originally developed to help organizations identify security vulnerabilities in their own systems through automated testing. According to analysis by Microsoft, the artificial intelligence agent within ARTEX examines system defenses, exploits vulnerabilities, and can extract data. Attackers leverage this automation to attempt network penetration through multiple methods simultaneously.

One documented case involved an attack on Korean financial institutions that resulted in the exposure of approximately 67,000 individuals' personal information. In another series of attacks, threat actors stole authentication credentials for ConnectWise backup software and used ARTEX alongside it to attempt sign-ins to more than 2,250 Microsoft 365 customer environments. Because customers had pre-authorized access to the backup software, attackers could exploit the permissions granted to the software rather than requiring individual user accounts. Microsoft noted that successful sign-in records alone do not confirm data access or exfiltration occurred.

Amazon Web Services released a research report on November 13 describing a technique where attackers steal artificial intelligence service credentials, use those services at others' expense, and search for additional credentials. Amazon Threat Intelligence, the company's threat analysis division, identified this activity through decoy systems posing as artificial intelligence services and assessed with moderate confidence that ARTEX was used in these attacks.

According to Amazon, attacks exploited default passwords, externally exposed login credentials, and unpatched software. The use of artificial intelligence enables attackers to identify and exploit these weaknesses in shorter timeframes. The Federal Bureau of Investigation stated that ARTEX's developers removed the original repository from GitHub in October following reports of misuse, though copies continue to circulate. The agency recommends restricting access to sensitive systems, reviewing software permissions, updating exposed credentials, and monitoring for suspicious account activity and large data transfers.