Home›Crypto›Market Structure›AI helps researcher craft working Zoom exploit in unde…

AI helps researcher craft working Zoom exploit in under 24 hours

The reported flaws, tracked as CVE-2026-53413, CVE-2026-53414, and CVE-2026-53415, target Zoom’s annotation tool and could let an attacker control a participant’s device without any action by the victim.

Reported · Written up by our desk · Market Structure

A Security reports that a researcher used publicly available AI tools to find and build a working exploit for Zoom vulnerabilities in less than 24 hours, using fewer than 20 AI prompts.

The firm says the bugs affect Zoom’s annotation tool and could allow an attacker to run code on another participant’s device without the victim clicking anything, potentially enabling actions like accessing personal data or turning on the microphone or camera.

According to A Security, the proof-of-concept was tested across Zoom apps for Windows, macOS, Linux, Android, and iOS. The company also says the attack could work both ways in a meeting, with a compromised presenter reaching participants and a compromised participant reaching the presenter.

A Security said it reported the first flaw to Zoom on June 10, and Zoom released fixes between June 22 and July 20. The report notes that even after fixes, users still needed to update because server-side safeguards could not filter malicious messages in end-to-end encrypted meetings, and a Zoom spokesperson told Decrypt that users should keep their apps up to date.

More like this

Sources