a bug called 'Zoomsday' let any meeting participant silently run code on your Mac or iPhone. researchers built the working exploit using an AI model and fewer than 20 prompts.
Zoom quietly patched four vulnerabilities on August 11, 2026 — and the worst one, CVE-2026-53413, needed nothing from you at all to get exploited. Nicknamed 'Zoomsday' by the researchers who found it, the bug lived in Zoom's annotation engine and let any other participant in a meeting silently execute code on your Mac, iPhone, Windows PC, Android device or Linux box. No link to click, no file to open, no permission prompt to approve — just joining a call with the wrong person in it was enough. Zoom has already shipped fixes. If you haven't updated your Zoom client this week, that's the one thing to do before you finish reading this.
Open Zoom, go to your profile picture, and click 'Check for Updates' — don't wait for auto-update to catch it.
Confirm you're on Zoom Workplace 7.1.5 (or 7.0.6) or newer — anything older is still exposed to the annotation-engine bugs.
If your organization runs Zoom Rooms or the Meeting SDK, confirm IT has pushed 7.1.5 to those deployments too — client-side updates alone don't cover them.
Running Zoom's VDI Client or VDI Plugin? Update to 7.0.11/6.6.16 (Client) or 7.0.11/6.6.15 (Plugin) separately — that's a different bug (CVE-2026-53416) with its own fix.
The mechanics of the attack are almost embarrassingly simple once you see them laid out. Zoom's in-call annotation tool — the pen and shape overlay you use to mark up a shared screen — didn't properly validate what it was being asked to draw. A malicious participant could craft an annotation payload that got interpreted as executable instructions instead of pixels, and because annotations render automatically for everyone in the call, the payload fired the instant it was sent. The target didn't need to be presenting, sharing their screen, or doing anything unusual — just sitting in the meeting.
How an AI cracked Zoom in under 20 prompts
The part of this story that should worry security teams more than the bug itself is how it was found. A Security cofounder Omer Gull put it bluntly: work that used to take a team of five researchers roughly six months now took under 20 prompts to a publicly available AI model, start to finish, in under a day. That's not a claim about one lucky bug — it's a claim about the entire economics of vulnerability discovery shifting underneath the software industry in real time. A well-resourced attacker doesn't need a dedicated exploit-dev team anymore to find something like this; they need an AI model, a target, and a weekend.
A Security says it built a full working exploit for the Zoom bug using publicly available AI models in under 24 hours. · Unsplash
The other three bugs nobody's talking about
All four disclosed Zoom vulnerabilities (Aug 11, 2026)
For most consumer and business users, updating your Zoom client to the latest version closes all four bugs. VDI Client/Plugin users need a separate, additional update — check your version against the table above.
Was this vulnerability actually exploited by real attackers before the patch?
Zoom and A Security have not reported evidence of in-the-wild exploitation before disclosure. This was found and reported through responsible research, not discovered as an active attack.
Does this affect the free version of Zoom too?
Yes — the annotation engine bug affects Zoom Workplace broadly, which includes free-tier accounts. Severity doesn't depend on your subscription plan.
Update Zoom today if you haven't — this isn't a 'patch when convenient' bug given the zero-click, zero-interaction nature of the exploit. Beyond that, keep an eye on whether A Security or other researchers publish more detail on the underlying annotation-parsing flaw once patch adoption is high enough that publishing exploit specifics won't hand attackers a roadmap to unpatched systems.
