Agentic AI Browsers: Hilariously Easy Targets for Scammers Everywhere!
Agentic AI browsers like Perplexity’s Comet are entering the mainstream but are fraught with security issues. Guardio discovered they can fall for phishing scams, fake shops, and prompt injection attacks. In one test, Comet bought an Apple watch from a fake site! Until security improves, keep sensitive tasks away from these AI agents.
Hot Take:
Agentic AI browsers are like toddlers let loose in a candy store—full of potential but disastrously unprepared for the world’s dangers. They’ll click anything shiny and autofill those credit card details faster than you can say “Oops!” Let’s just say, when it comes to security, these AI browsers are still in their digital diapers.
Key Points:
- Agentic AI browsers can autonomously perform online tasks but are vulnerable to both old and novel cyber threats.
- Comet, a key player in this field, has already ventured into mainstream consumer markets.
- Security tests reveal these browsers easily fall for phishing schemes, prompt injections, and fake websites.
- Guardio’s research highlights the risk of AI models being exploited once, then scaled for endless scams.
- Users are advised against delegating sensitive tasks to these AI browsers until security measures improve.
Already a member? Log in here