Hackers exploit AI chatbots to spread malware via search results
Researchers discovered that attackers exploit AI chatbots to seed Google search results with malicious terminal commands. When users search common topics and click sponsored links to ChatGPT or Grok, they unknowingly execute commands that grant hackers system access to install malware like AMOS.
What happened
Attackers prompt AI chatbots ChatGPT and Grok to suggest pasting harmful commands into computer terminals during conversations about common search terms. The hackers make these conversations publicly visible and pay to boost them on Google Search.
When users search for the topic, the malicious AI chat links appear high in search results. Users unknowingly click the sponsored links and execute the commands, granting attackers system access to install malware.
Reported impact
A real-world attack used this method to target Mac users searching for 'clear disk space on Mac.' Users clicked a sponsored ChatGPT link and executed a command that installed the AMOS malware, a Mac-targeting data exfiltration tool.
The attack bypasses traditional security warnings because users trust Google and ChatGPT. No file download, suspicious executable, or obviously shady link is required to compromise the device.
Index analysis
Huntress testing confirmed that both ChatGPT and Grok can be manipulated to suggest harmful terminal commands in chat conversations that appear in Google search results. A malicious ChatGPT link remained visible in Google search results for at least half a day after the security firm's disclosure.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.