How was the Zoom vulnerability discovered?

Researchers at A Security discovered the flaw using fewer than 20 prompts on publicly available AI models.

What feature in Zoom did the exploit target?

The exploit targeted Zoom's annotation feature used during screen sharing.

What could an attacker do by exploiting this Zoom flaw?

An attacker could run malicious code to steal data, turn on the camera or microphone, or install malware on a victim's device.

Was any user action required for this Zoom attack to work?

No, the attack required no action from victims beyond their participation in the compromised meeting.

Has the Zoom vulnerability been fixed?

Yes, Zoom has issued a patch for this security vulnerability.

AI Wrote a Zoom Hack in Under 20 Prompts

The patch is in place, but the implications are not. This event, dubbed "Zoomsday," signals that the barrier for sophisticated, automated cyberattacks has just collapsed. Where a nation-state team with months of effort was once required, now a single researcher with an AI agent and one day of work can do the same.

Why your next Zoom meeting could be an open door to hackers

For millions of remote workers and teams, the daily Zoom call is a mundane necessity. The "Zoomsday" hack reveals it can also be a total compromise vector. This isn't a phishing email requiring a user to click a link. It's a flaw that could be exploited during any session where screen sharing is used.

The scariest part of the story for non-technical users is the "fewer than 20 AI prompts" claim. It crystallizes a new threat paradigm: the weaponization of productivity features is now scalable through automation. The features designed for collaboration, like screen sharing with annotations, are being reverse-engineered by AI into attack surfaces. And there are no visual cues to warn you.

This moves the risk from user error to systemic vulnerability. As we reported in Coldcard Hack Voids $100 Million Self-Custody Promise, foundational security assumptions in widely trusted tools are being shattered. With Zoom, the assumption was that joining a meeting couldn't lead to complete device hijacking. That assumption is now incorrect.

How a simple annotation tool became a weapon

Zoom's annotation feature is a staple of collaborative calls. When a participant shares their screen, others can draw, highlight, or add text directly onto the shared view. It's meant for pointing out details, not for executing code.

The exploit, discovered by researchers at A Security, twisted this function into an attack channel. While the source material does not detail the exact technical mechanism, the outcome is clear: an attacker, by joining or hosting a meeting, could leverage the annotation feature to run malicious code on victims' devices.

The attack required no action from the victims beyond their participation in the compromised meeting. Critically, A Security's blog post notes it showed "no visual cue indicating the compromise." Your screen would look normal. Meanwhile, in the background, your device could be wide open.

The AI accomplice: Automating a sophisticated attack

Traditionally, finding and exploiting a complex vulnerability like this is described as "nation-state work: elite teams, months of effort, budgets that governments regulate as weapons," as A Security researcher Idan Levcovich wrote. Here, they did it in a single day with an AI agent.

XOOMAR Inference: While the source doesn't list the prompts, we can infer the AI's role from the outcome. The AI models likely accelerated two key phases:

Discovery: Guiding the AI to analyze Zoom's attack surface, specifically focusing on real-time collaboration features like screen sharing and annotation, which handle complex, bidirectional data.
Exploitation: Helping craft the malicious payload that could be delivered through the annotation data stream, turning a display feature into a code execution engine.

This shift is profound. Manual security research is a grind of intuition and deep expertise. AI-powered research can systematically probe thousands of potential attack vectors in hours, drastically shortening the time from "potential flaw" to "working weapon." It democratizes high-end offensive capabilities.

From stealing data to turning on your webcam: The hacker's playbook

Once the exploit was successful, the potential for harm was virtually unlimited. According to the researchers, the malicious code could enable an attacker to:

Steal data from the compromised device.
Turn on the camera or microphone for surveillance.
Install malware for persistent access.

Consider the settings where this would be catastrophic. A board meeting discussing an unannounced acquisition. A healthcare provider reviewing patient records. A lawyer walking a client through sensitive documents. The annotation tool, meant to clarify, could instead be the silent conduit for total data exfiltration or espionage. The stealth is what makes it so powerful, no pop-ups, no warnings, just a normal meeting.

Is your Zoom app patched, and what comes after Zoomsday?

Zoom issued a fix for the vulnerability on Tuesday, August 11, 2026. The patch impacts the app across Windows, macOS, Linux, Android, and iOS. Your immediate action is simple:

Action: Force-quit and restart your Zoom desktop or mobile app. This should trigger an update check. Ensure you are running the latest version. The patch is likely automatic for most users, but a restart guarantees it.

The bigger question is whether this is an isolated flaw or a symptom of a deeper issue. As collaboration tools bundle ever more real-time, interactive features (co-editing, whiteboards, live reactions), their attack surface expands exponentially. This incident proves that AI can now probe that surface with terrifying efficiency.

What to watch: The cybersecurity industry's response. If AI can crack a major app's core security in under 20 tries, defensive tools must also evolve at AI-native speed. We are moving from an era of human versus human in cybersecurity to AI versus AI. Defenders will need to deploy their own AI agents to continuously audit code and monitor for novel attack patterns, a theme highlighted in our coverage of Safety Tests Unleash AI Agents That Hack Production Systems.

Zoomsday is a warning shot. The next target could be any platform where we work, meet, and share in real time. Your best defense, for now, remains the mundane but critical habit of keeping every single app updated, immediately.

Impact Analysis

The barrier for sophisticated cyberattacks has collapsed, moving from nation-state teams requiring months of effort to a single researcher with an AI agent needing only one day.
This event transforms widely used productivity features, like screen sharing and annotations, into scalable attack surfaces with no visual warnings for users.
Foundational security assumptions for millions of remote workers are shattered, as merely joining a Zoom meeting can now potentially lead to complete device hijacking.