Trick Perplexity's Comet AI Browser Into Phishing Scam in Under Four Minutes

Guardio reports that it tricked Perplexity's Comet AI browser into falling for a phishing scam in under four minutes by exploiting the agent's tendency to narrate its reasoning. Guardio said it intercepted traffic between the browser and the vendor's AI services and fed it into a Generative Adversarial Network, which iteratively regenerated a phishing page until the browser proceeded to enter credentials on a bogus refund page. Security researcher Shaked Chen said the same approach could work on every user relying on that agent, since the target shifts from the human to the AI. The article also cites separate Trail of Bits and Zenity Labs research on prompt injection flaws in Comet, which Perplexity has addressed, and notes OpenAI's view that such weaknesses are "unlikely to ever" be fully resolved in agentic browsers.