On November 9, 2026, Telegram quietly opened a new door with Bot API 10.4—developers can directly embed interactive applications within chat messages, executed by QuickJS to handle script logic, while there is currently no clear information publicly available regarding sandbox isolation. Three days later, the founder of SlowMist, Yuxian, issued a rare warning, specifically mentioning that this new entry point could lead to risks of malicious code execution and sandbox escape, even suggesting that regular users should delay updates or limit interaction with these "flashy" new features. Almost simultaneously, Lorenzo Valente, head of digital assets research at ARK Invest, provided a chilling industry summary: since 2017, the crypto world has suffered approximately $26.9 billion in losses due to hacker attacks, while on-chain revenue during the same period was about $47 billion. Statistically, this translates to approximately $0.57 being hacked for every $1 of on-chain revenue. The closer the entry point is to users and the smoother the experience, the more willing attackers are to exploit these gaps. With AI significantly enhancing the efficiency of vulnerability digging and exploitation, the tug-of-war around "experience innovation" and "security costs" is becoming exceptionally clear at this newly opened front-end entry of Telegram.

The New Telegram Feature Transforming Chat Windows into App Stores

On November 9, 2026, Telegram released Bot API 10.4, elevating a change that originally belonged to "underlying interface updates" to a closer user entry point—developers can directly embed interactive "applications" within a single message, executed by the client in the chat window using JavaScript logic. For regular users, this means that the dialog box is no longer just text and buttons; it can be "refreshed" at any time into a mini application interface that is clickable, input-capable, and computational. For developers, this equates to having an additional “advertising space” for real-time code distribution and execution within each message. Given Telegram's already dense network of crypto project communities and bot ecosystems, many on-chain related operations that were previously completed via link redirects or command-line interactions now have the opportunity to be wrapped as innocuous-looking chat cards that can be "sorted out with just a click" within conversations.

However, risks have quietly embedded themselves within this smooth experience: these embedded applications execute scripts on the client side, running logic through the QuickJS engine, while currently public information does not reflect that strict sandbox isolation mechanisms have been activated on the client side. For long-term users who treat Telegram as a bulletin board or group chat tool, the same chat window serves as both message flow and code execution environment, blurring the boundaries between information and programs. Yet, the vast majority cannot see the script content and are unable to determine whether it has been properly isolated. The feature is still in the early promotional phase, with its interface closely resembling traditional bot messages, leading users to naturally carry over past habits of "clicking when seen and trusting familiar interfaces," treating it as an ordinary interaction rather than a potential local code execution. This is fundamentally why it is regarded as a new risk entry point.

QuickJS and the Line of Malicious Code Lacking a Sandbox

According to public information, the interactive applications embedded in Telegram's client are executed through the QuickJS JavaScript engine. The issue is not about whether the QuickJS implementation is "good" or "bad," but rather what kind of container it has been placed in: Yuxian pointed out that this implementation lacks clear sandbox protection boundaries, and the isolation layer between the script and the host environment is not transparent. The path for malicious code to trigger execution indeed exists, and once coupled with implementation flaws in the engine or integration layer, there is potential for sandbox escape. For regular users, this often amounts to an invisible line of script, but for attackers, it could be a "remote switch" attached to a message, especially as AI-assisted vulnerability digging and crafting exploitation chains continues to lower barriers. As long as there is an entry point, the cost of attempting to pry open access becomes unprecedentedly low.

Consequently, Yuxian publicly advised on November 12 that regular users at this stage can delay updates or minimize use of such "flashy" new features, focusing their interactions on the functionalities they genuinely need. For developers, they should treat this entry as a high-risk interface and avoid encouraging users to frequently click on complex script-driven interactions in their application design. It is important to emphasize that, to date, there have been no publicly disclosed cases of actual attacks utilizing this embedded application feature; the risk remains at an early warning level on a technical front. However, given the lack of complete sandbox disclosures and an official repair timeline, it is more fitting to view it as a new attack surface that warrants additional vigilance rather than a readily usable everyday feature.

$26.9 Billion Stolen Compared to $47 Billion in Revenue

If we place the technical warning regarding Telegram's new entry point on a longer timeline, it is not facing abstract "risks," but rather a series of quantifiable and remarkably heavy historical losses. Lorenzo Valente from ARK Invest revealed that since 2017, the crypto industry has incurred cumulative losses of approximately $26.9 billion due to hacker attacks; during the same period, total industry on-chain revenue was about $47 billion. In intuitive proportional terms, under his statistical criteria, every $1 of on-chain revenue corresponds to approximately $0.57 being taken by hackers—meaning, the safety gaps have long eroded more than half of the industry's economic gains. More extreme tail events are not entirely absent, with one source reporting a single hacker attack against Binance that caused losses of approximately $6.36 billion, allegedly exceeding the entire 2026 on-chain revenue. This figure and its comparative conclusions currently have moderate confidence and should be approached with caution, but even taking it as a potential scale reference underscores that the cost of compromising a single entry point can be substantial enough to "wipe out a year’s accounts."

Against this backdrop, Telegram executing JavaScript directly within the chat window without publicly reflecting a designed strict sandbox isolation is no longer simply "a certain application adding a whimsical feature," but is naturally classified by security practitioners as yet another high-risk entry. Over the past decade, the $26.9 billion in stolen funds did not result from a few random incidents but from a cycle of different entry points and attack surfaces being repeatedly explored and exploited. Today, with the capabilities of AI-assisted vulnerability digging and exploitation noticeably enhanced compared to the past, any new interface lacking isolation theoretically lowers the barriers to attack even further. In light of the long-term comparison of $0.57 losses versus $1 revenue, treating Telegram's new feature as a structural security issue rather than a one-time "misupdate" is a more responsible interpretation of these historical figures.

When Telegram Chats Become a Part of the Attack Landscape

In the crypto world, Telegram has long ceased to be merely a "chat software" but has transformed into a hybrid of bulletin board, customer service desk, and trading front-end: projects release white paper updates here, communities participate in whitelist races, and users habitually check balances and place orders via bots. Following Bot API 10.4, these actions began to be further folded into a single chat message—embedded applications directly executing JavaScript within the chat window allow users to complete interactions without navigating to web pages or separately opening mini-programs. The entire interaction is compressed into a smooth path of "clicking the message—performing the operation within the bubble." It is precisely this "seamless connection" that quietly integrates the chat window into the list of options for attack path design.

For attackers, the script thus becomes smoother: first, through community operation or disguising official channels, keeping target users within Telegram chats, then replying or announcing messages with seemingly normal bots containing embedded application entrances, leading users to authorize, sign, or bind accounts within familiar dialog bubbles. A single message can carry both social engineering (identity spoofing, dialogue) and executable scripts (QuickJS running logic), weaving together chat entry, phishing links, malicious bots, and subsequent wallet interactions into a nearly undetectable "cross-scene" attack chain. One of Yuxian's points of caution is here: at a time when AI has significantly lowered the barriers to vulnerability mining and exploitation, any new entry lacking clear separation becomes exponentially more vital to security once it connects organically to users' financial operation paths. Although as of November 12, 2026, there has been no public evidence demonstrating that this new feature has been utilized in large-scale practical applications, nor has Telegram provided thorough responses or repair timelines regarding specific technical concerns, when the chat window itself becomes a potential execution environment, it is no longer "the front end for observing market trends" but rather an infrastructure to be scrutinized as a potential attack node.

Upgraded AI Attack and Defense in the Context of Telegram Risks and Self-Rescue

Looking back at the approximately $26.9 billion in stolen losses from this single entry point since 2017, contrasted with roughly $47 billion in on-chain revenue during the same period, security costs can no longer be seen as lessons from only a few projects but rather as the lifeline for the entire industry. When the chat window is endowed with the capability to execute JavaScript while lacking clear sandbox isolation as indicated by public information, it effectively moves to the front line of the AI arms race: attackers can use AI to identify vulnerabilities and write exploitation chains more swiftly, while defenders are also forced to use AI for anomaly detection and rapid blocking, making any "flashy" new interactions potentially amplify risks in this round of the arms race. Yuxian's reminders regarding Telegram's new feature fundamentally question a long-accepted design logic—while the industry pursues "zero-friction entry" and high-frequency interactions, it often places security after user experience. Now, modeling threats, enforcing least privilege, and fostering cross-ecosystem collaboration must become preconditions in the product development phase. Currently, there is no authoritative data indicating that Telegram’s new feature has triggered a systemic event, yet discussions surrounding entry security and default trust boundaries are evidently heating up. What truly warrants attention moving forward is whether Telegram will strengthen its sandbox isolation and permission constraints and whether mainstream wallets and trading tools will lower their default trust levels for chat entries, introducing more granular interaction isolation mechanisms. These specific product boundary adjustments will decide whether Telegram-style entry points become a turning point in security or the starting point of the next loss curve.

Join our community, let’s discuss together and grow stronger!
AiCoin Exclusive Hyperliquid Benefits: https://app.hyperliquid.xyz/join/AICOIN88
AiCoin Exclusive Aster Benefits: https://www.asterdex.com/zh-CN/referral/9C50e2
On-chain Telegram Community: https://t.me/AiCoinWhaleData
On-chain Community: https://www.aicoin.com/link/chat?cid=N6OVMor5g
AiCoin On-chain Twitter: https://x.com/aicoinwhaledata

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到[email protected]，本平台相关工作人员将会进行核查。