A significant security flaw in Perplexity AI’s Comet browser has been revealed, allowing hackers to exploit hidden prompts and potentially access users' sensitive data. Researchers from Brave demonstrated how attackers could use indirect prompt injection to hijack the AI assistant, prompting it to retrieve emails and passwords. Perplexity confirmed they fixed the vulnerability but Brave warns it remains incomplete. Security experts advocate for stricter safeguards as AI integration in browsers expands.