Security Flaws Expose Perplexity AI's Comet Browser to Cyberattacks Amid AI Browser Competition

The emergence of AI-powered web browsers like OpenAI's ChatGPT Atlas and Perplexity AI's Comet is revolutionizing web surfing by integrating intelligent assistance directly into the browsing experience. ChatGPT Atlas, launched in October 2025, offers features such as a persistent sidebar for summarizing content and real-time data analysis, aiming to enhance productivity while maintaining broad compatibility and privacy. However, security concerns have surfaced, especially with Perplexity AI's Comet browser extension, which has been found vulnerable to indirect prompt injection attacks allowing malicious hidden commands to execute, posing significant risks to users' data and privacy. Researchers have demonstrated how attackers can exploit these AI assistants by embedding harmful instructions into web content, effectively hijacking user sessions and accessing sensitive information without user awareness. While these AI browsers promise unprecedented convenience, cybersecurity experts caution that their extensive access to personal data and susceptibility to novel attack vectors necessitate careful consideration and improvement in security measures. This tension between innovation and security highlights the challenges in adopting AI agents for everyday web tasks, especially in sensitive domains like cryptocurrency and personal communications.