Thought Leadership
Security researchers at Zscaler have uncovered a new attack campaign in which cybercriminals use manipulated AI chats to distribute malware.
The campaign primarily targets software developers and IT professionals, tricking them into executing malicious commands through seemingly trustworthy content.
According to the threat hunting team, the campaign combines two established techniques: the social engineering tactic known as “ClickFix” and Anthropic’s legitimate feature for sharing conversations from its Claude AI assistant.
The attackers create maliciously crafted, publicly shareable AI chats that appear authentic at first glance. Users are directed to this content through channels including online advertisements and are presented with instructions that allegedly are required to install an update, set up a development environment, or complete a security check.
Malware Is Launched Manually
The attackers’ main objective is to persuade victims to manually paste preconfigured commands into the macOS Terminal or Windows Command Prompt. Known as “ClickFix” or “InstallFix,” the technique bypasses traditional security mechanisms because the code is executed directly by the user.
Once the first command is entered, a multi-stage infection chain begins. Additional malware is downloaded in the background, often without the victim immediately realizing that an attack is underway.
The campaign specifically targets developer workstations. After infecting a device, the malware attempts to harvest a wide range of sensitive information, including browser credentials, session cookies, stored cryptocurrency wallets, and various system details.
Stolen login credentials and session tokens are particularly dangerous. If attackers gain access to this information, they can use it to move laterally across corporate networks and compromise additional systems.
Malware Adapts to New Attack Opportunities
Zscaler’s analysis shows that the attackers modified their tactics several times during the campaign. Rather than relying on fake websites, they are now abusing legitimate features of a well-known AI platform. This makes the malicious content appear more credible and increases the likelihood that users will follow the embedded instructions.
The campaign highlights how cybercriminals are increasingly incorporating popular AI services into their attack strategies and deliberately exploiting users’ trust in these platforms.
Security Awareness Remains Critical
Although the observed campaign was active for only a limited period, security researchers see it as a sign of a broader and growing trend. Organizations should therefore train employees never to blindly copy commands from unknown or unverified sources into terminal or command-line windows.
Developers and IT administrators are particularly attractive targets because they regularly work with scripts and console commands. A combination of security awareness, clear policies, and technical safeguards can help significantly reduce the risk of successful attacks.
(pd/Zscaler)
