By skyforbes So I've been playing around with a few of the new AI browsers everyone's talking about, and… yeah, not all of them seem ready for prime time.
Turns out one of them (Comet) has a pretty nasty unpatched vulnerability that can literally grab your emails through a crafted URL. It hides prompts in Base64 and slips past the usual protections.
Then there's Atlas, which apparently relies on users to "just keep an eye" on what the agent mode is doing. Sounds super safe, right?
Feels like this whole "AI browser" thing is moving way faster than the security side can keep up. Anyone else tried them yet? What's your take so far?