🤖 AI found a Zoom vulnerability before any human did

So researchers from A Security just disclosed some pretty serious Zoom flaws — and the wild part is how they found them. Using publicly available AI models, fewer than 20 prompts, they had a working exploit. Their cofounder told WIRED that the same work would've taken a five-person team about six months before.

The bug was in the annotation protocol used during screen sharing. Anyone on a call — host, participants, didn't matter — could've been silently compromised with zero interaction from their side. All major platforms: Windows, macOS, Linux, iOS, Android.

Zoom's already rolling out fixes. But the barrier-to-entry thing is what's actually unsettling here — that's not going back up.

Ars Technica