Images from AI chatbots may contain hidden malware. New research reveals how LLMs can be manipulated through image scaling vulnerabilities.

Researchers reveal how downscaled images can hide malicious prompts, raising risks of identity theft and sensitive data exfiltration.