Crashing Comet: Inside the Hunt for AI Browser Vulnerabilities with CyberArk Labs

AI powered browsers are reshaping how users interact with the web, but new research shows they may also be introducing serious security gaps. This session takes a close look at how CyberArk Labs uncovered critical vulnerabilities in Perplexity’s Comet browser, including prompt injection attacks that enabled unauthorized purchases, local file access and full account takeovers. The discussion explains how these weaknesses challenge long standing browser security assumptions.

Researchers walk through the discovery process, from early testing to coordinated disclosure, offering practical insight into how modern vulnerability research unfolds. The webinar highlights why large language models cannot be treated as security boundaries and what these findings mean for organizations evaluating AI driven tools. Attendees will gain a clearer understanding of the risks surrounding AI integrated browsers and the broader implications for future web security models.