Sunday, 26 October, 2025
The Security Threat of AI Browser Agents
New AI browser agents, like those from OpenAI and Perplexity, present major security and privacy risks. These agents, designed to automate online tasks, can be tricked by hidden malicious instructions—a "prompt injection attack"—into performing actions not intended by the user, such as leaking sensitive data, posting to social media, or executing unauthorized purchases. Experts caution that current fixes are insufficient.
Read full story at TechCrunch