As AI models gain advanced capabilities to find vulnerabilities in software, develop ways to exploit them, and even carry out autonomous hacking sprees, researchers offered a sobering new example on Tuesday by disclosing vulnerabilities in the video conferencing platform Zoom that could have been exploited to take over targets’ devices. Anyone on a call that involved screen sharing, whether participants or the host, would have been vulnerable to a silent attack that could be carried out with no indication and no interaction from the victim. This Zoom screen-sharing bug lets attackers take over devices simply by getting a target onto a shared screen call, underscoring a dangerous shift in the cybersecurity landscape.
AI-Powered Vulnerability Discovery: Under 20 Prompts to Weaponize a Zero-Day
Researchers from the digital defense firm A Security discovered the bug in early June using publicly available AI models, requiring fewer than 20 prompts to uncover the vulnerabilities and create a working attack. The flaws affected devices running all operating systems Zoom supports — Windows, macOS, Linux, iOS, and Android. Zoom issued a security advisory on Tuesday, including details about fixes the company has already begun rolling out.
How Did AI Models Discover the Zoom Screen-Sharing Vulnerability?
The AI bug hunting systems used by A Security specifically targeted the real-time annotation protocol during screen sharing. Because the component is convoluted, obscure, and part of proprietary closed-source software, the AI models were trained to prioritize such functions — exactly as human bug hunters would. Within 20 prompts, the AI identified exploitable weaknesses that would have taken a team of five people six months to find using traditional methods.
“What is interesting for us and what we believe is dangerous is the democratization of these capabilities — the barrier to entry is dropping rapidly,” A Security cofounder Omer Gull said ahead of the disclosure. “Before it would have taken a team of five people maybe six months with a lot of refining and iteration to find this. Now people can reach the same results with under 20 prompts. And Zoom is an important type of target because people assume trust when using it. They don’t see it as a threat.”
The Annotation Protocol: Why Closed-Source Features Breed Hidden Flaws
The vulnerabilities were specifically in the protocol used to facilitate real-time annotation during screen sharing. The researchers explained that their AI bug hunting systems delved into this component because convoluted and obscure functions often contain overlooked vulnerabilities — a truism especially valid for proprietary, closed-source software. While an established company like Zoom presumably does extensive code review and vetting on all components, without the benefit of public, open review, esoteric yet complex features like annotation are more likely to contain mistakes.
A Security cofounder Yossi Torati described the chilling simplicity of the attack: “If you just get on a Zoom with us, we can take over your device.” He added, “The worst-case scenario is that we can take over an enterprise just by having this vulnerability in our hands. If I’m an attacker, I can be on a call with someone from a company, take control of their computer and their credentials, and then use them to move laterally in the enterprise.”
Patching the Bug: Server and Client-Side Fixes Now Deployed
Zoom has now patched the vulnerabilities, issuing both server and client-side fixes — updates for Zoom’s own servers and the applications running on customer devices. The company did not respond to multiple requests for comment about the A Security findings, but the security advisory details the fixes that address the flaws. The researchers emphasized that it was alarming to contemplate bugs that could have been exploited to take over a target device simply by getting someone onto a Zoom call. Joining a call is in itself a gesture of trust, but given how ubiquitous video calling is in personal and professional contexts — and given that Zoom in particular is widely used for events and semipublic activities like webinars — people typically have their guard down when joining.
The Democratization of Offensive Cybersecurity: AI Lowers the Barrier to Entry
Practitioners often call security a “cat-and-mouse game,” but as AI bug hunting proliferates, this delicate dance has become an all-out race. The ability to weaponize a zero-day with fewer than 20 AI prompts is not just a technical milestone — it represents a fundamental shift in who can become a threat actor. Historically, discovering and exploiting vulnerabilities in enterprise-grade software required deep expertise, expensive tooling, and months of manual effort. Now publicly available AI models can achieve comparable results in hours or days, putting sophisticated attack capabilities in the hands of a far wider pool of adversaries.
This democratization has profound implications for enterprises. The attack vector described — silent, requiring no victim interaction, and applicable to any operating system — means that a single compromised screen-sharing session could cascade into a full-scale network breach. Credentials harvested from a host’s device could enable lateral movement across an organization, exposing sensitive data, internal systems, and interconnected partners.
What Does This Mean for Video Conferencing Security Going Forward?
The Zoom case is a harbinger of a broader trend. As AI models become more adept at analyzing closed-source software, the number of zero-day discoveries will likely accelerate. Developers will face mounting pressure to harden every obscure feature of their applications, because AI-driven bug hunters will systematically probe them. For video conferencing platforms — a category that relies on deep user trust and operates across personal, corporate, and public contexts — the risk is especially acute. Features like annotation, chat, file sharing, and screen recording are fertile ground for hidden flaws.
For users, the lesson is clear: joining a video call with screen sharing enabled should no longer be considered a low-risk activity. Even patched vulnerabilities leave a residue of concern, as similar bugs in other components may still exist. Enterprises should enforce strict policies on who can share screens during internal or external meetings, monitor for unusual behavior during calls, and ensure that endpoint protection systems are updated to detect exploitation attempts targeting video conferencing software.
The race between AI-powered discovery and automated patch deployment is intensifying. Zoom’s rapid response — issuing both server and client-side fixes — demonstrates that detection speed has improved, but the window of exposure remains dangerously short. As A Security’s findings show, the next zero-day may be uncovered in even fewer prompts, and the target may not be a video conferencing tool but a core enterprise system, a critical infrastructure component, or a widely used consumer app. The cat-and-mouse game has become a sprint, and the mice are now running on AI-powered legs.
