Security researcher Gal Weizman disclosed BragJack, an attack technique that hijacks AI browser assistants through malicious extensions. The proof-of-concept targeted five browsers including Google Chrome's Gemini Live, Microsoft Edge, Perplexity Comet, Opera Neon, and Anthropic's Claude, earning over $20,000 in bug bounties and producing two CVEs. The attacks exploit how AI assistants are integrated into browsers by manipulating web traffic through Chromium's declarativeNetRequest functionality to intercept requests and gain unauthorized access to sensitive data or control agent actions.