Google discovers malware using AI to evade detection
Google reported discovering three malware strains that connect to large language models including Gemini to refine attacks in real-time. The malware Promptflux reportedly queries Gemini to generate code that evades antivirus detection, while Quietvault uses AI to locate and steal secrets from infected systems.
What happened
Google's Threat Intelligence Group warned in a November 5, 2025 report that three malware strains were found connecting to large language models and being used in actual operations.
Promptflux malware queried Google's Gemini chatbot to modify its code and avoid antivirus detection, using prompts such as 'Provide a single, small, self-contained VBScript function or code block that helps evade antivirus detection.' Quietvault malware used AI prompts and on-host installed AI tools to search for secrets and exfiltrate files from infected Windows systems.
Reported impact
Google stated that Promptflux could rewrite the malware's entire source code on an hourly basis to evade detection.
Index analysis
Security researcher Marcus Hutchins questioned the threat posed by the AI-generated malware, citing weak or impractical prompts and stating that Gemini did not instinctively know how to evade antivirus software.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.