Security researchers at NebuSec have reported that a malicious widget can successfully bypass the security filter recently introduced by Telegram to its Android application, allowing unauthorized code to execute outside of designated security boundaries.

The security concern revolves around interactive widgets—small interactive blocks that bots can send within chat messages, executing JavaScript directly on the user’s mobile device. Because Telegram currently restricts live widget deployment exclusively to approved bots, the research team conducted their recent demonstration within a controlled Telegram test server environment.

The chain of events leading to these findings began shortly after Telegram enabled bots to start sending widgets. Just two days later, an account identified as V12 posted a video on the social media platform X, claiming that arbitrary Telegram messages could be leveraged to achieve remote code execution. In response to these early reports, Telegram deployed updates in version 13.2.2 (7303) of its Android application. These updates isolated QuickJS, the lightweight engine responsible for running widget JavaScript, into a dedicated process. Additionally, the developers introduced a seccomp filter designed to intercept, record, or block specific system calls made by that isolated process.

However, in a technical analysis published by NebuSec, researchers revealed that they successfully bypassed this filter using two distinct methodologies. One method exploits a vulnerability within the widget handling mechanism that the filter fails to cover, while the second approach chains a covered widget bug alongside a kernel vulnerability. In a video demonstration shared alongside their findings, a bot-generated Telegram widget running on Android version 13.2.2 (7303) managed to circumvent the seccomp filter and automatically launch the device’s native calculator application.

Researchers Claim Two RCE Exploits That Bypass Telegram’s New Android Defenses

While the initial technical write-up and demonstration post on X did not disclose the full extent of what the executed code might be capable of reaching—such as accessing application messages or broader device storage—the lack of comprehensive public disclosures highlights ongoing security challenges. Furthermore, the public documentation does not specify whether user interaction, such as starting a bot or tapping a widget, is strictly required to trigger the behavior, nor does it assign a Common Vulnerabilities and Exposures (CVE) identifier or confirm whether the findings were formally reported to Telegram’s development team prior to publication.

According to Telegram’s official widget documentation, widgets are ostensibly designed to operate within a strict sandbox environment subject to restrictive rules, including a prohibition on network access. These restrictions are natively enforced within the engine’s internal codebase. Nevertheless, security experts note that a single memory corruption flaw, such as a use-after-free or an out-of-bounds write within the QuickJS engine, allows a malicious script to read and write directly to the process’s memory space, effectively neutralizing these internal sandbox boundaries.

The newly implemented seccomp filter explicitly targets a few dozen system calls—such as ptrace, bpf, and mount—while permitting all other system calls to pass through unchecked. An examination of the filter’s source code within the official Telegram Android repository indicates that the ioctl system call remains unfiltered. This filter operates as an auxiliary layer layered on top of existing Android-level limitations governing the engine process. Moreover, the architecture relies on a single shared process to handle all widgets, meaning that a compromised widget originating from a single bot shares memory resources with widgets deployed by any other active bot.

Code analysis conducted by industry researchers revealed that within the published source code for Telegram for Android version 13.2.2 (7303), the security filter was configured to operate in an audit mode. While operating in audit mode, the filter continues to block attempts to initiate new programs or processes, but it merely logs the remaining listed system calls rather than actively blocking them. Code comments left in the repository outline the intended duration of this monitoring phase, stating that the audit mode would remain active "until the logs from the devices say the engine never makes them." It remains unclear from the publicly available source code whether the distribution builds installed directly by end users operate under identical audit settings.

Researchers Claim Two RCE Exploits That Bypass Telegram’s New Android Defenses

Concurrently, community discussions on GitHub repositories dedicated to the Telegram bot API have addressed ongoing restrictions. A project contributor noted that general bots are currently prohibited from utilizing widgets, even within private chat interfaces. As of this week, no official updates have been issued indicating a relaxation of these restrictions or a specific timeline for deploying a permanent fix.

Addressing broader security concerns earlier, Telegram Chief Executive Pavel Durov stated in a public channel post that users could not be compromised simply by receiving an incoming message. His remarks addressed earlier reports regarding remote code execution via widgets across both Android and Windows platforms, emphasizing that successful exploitation required a user to actively initiate a third-party service, and asserting that the reported vector had already been fully patched on the server side. Telegram previously confirmed that the initial vulnerability highlighted in the V12 demonstration had been addressed and rolled out to users.

Despite these assurances, independent cybersecurity vendors have reported active exploitation attempts in the wild. Cybersecurity firm Kaspersky revealed that it detected suspected exploitation targeting a vulnerability within Telegram widgets while users browsed specific Telegram channels. In these observed instances, a PowerShell script executed silently in the background, subsequently downloading and launching a payload known as WebRAT, or Salat Stealer. This malware grants unauthorized attackers remote control over the infected system while harvesting sensitive data, including cryptocurrency wallet credentials and Telegram account information.

Kaspersky’s advisory did not specify the exact vulnerability exploited, the precise Telegram version targeted, or the total number of affected victims, nor did it directly link the activity to the recent researcher demonstrations. Security analysts continue to advise caution and recommend disabling widget features until an official, comprehensive software patch is released.