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

Hold up — AI browsers are getting played. Researchers just showed how to trick Perplexity’s Comet AI browser into a phishing scam in under four minutes. Yeah, you read that right. Four minutes.

Using GAN-trained phishing pages, they exposed a whole new attack surface that targets AI directly. This isn’t your grandma’s phishing — it’s next-level social engineering that bypasses traditional detection.

The method? Prompt injection + machine learning. They crafted pages that look legit to AI but are pure traps. Once the AI takes the bait, it can leak sensitive data or execute malicious actions.

This is a wake-up call for AI browser security. If AI can be fooled this fast, what’s stopping bad actors from scaling these attacks? The research highlights critical vulnerabilities in how AI interprets and interacts with web content.

Key takeaway: AI isn’t immune to old-school tricks — it’s just a new vector. As AI browsers become mainstream, securing them against these sophisticated attacks is non-negotiable.