All About Cookies is an independent, advertising-supported website. Some of the offers that appear on this site are from third-party advertisers from which All About Cookies receives compensation. This compensation may impact how and where products appear on this site (including, for example, the order in which they appear).
All About Cookies does not include all financial or credit offers that might be available to consumers nor do we include all companies or all available products. Information is accurate as of the publishing date and has not been provided or endorsed by the advertiser.
The All About Cookies editorial team strives to provide accurate, in-depth information and reviews to help you, our reader, make online privacy decisions with confidence. Here's what you can expect from us:
AI is making sophisticated cyberattacks easier to pull off, and a new WeChat worm shows just how low the technical barrier has become.
Security firm Calif has built an AI-assisted worm that can compromise WeChat accounts through an incoming call, without the victim answering or even touching their phone.[1]
Once compromised, the attacker can read and send messages in your name, make calls, and potentially access the payments and other services tied to your account.
The worm can then use the compromised account to target other people in the victim’s contact list, allowing the attack to spread from one account to another.
Security research firm Calif has built a zero-click worm known as WeWorm that can spread through WeChat calls and compromise your WeChat account. The victim doesn’t have to do anything for the attack to work.
The victim receives a WeChat call, but they don’t necessarily need to answer it for their account to be compromised. Even if they pick up, they hear nothing, and the exploit still succeeds. Declining the call stops that particular attempt, but the attacker can simply call again later, for example, when you are asleep.
Calif demonstrated the attack chain using three test phones. The first Android phone was the attacker, which called an iPhone and took over its WeChat while the call was still ringing.
The compromised iPhone was then used to call a second Android phone and take that over the same way. Accounts can be compromised within seconds and can allow the attacker to read and send messages through WeChat, make calls, and act as the account owner.
One prerequisite is that the threat actor must be on the contact list of the victim. However, that’s not necessarily a barrier, since an attacker can compromise one of your contacts first and then use their WeChat account to reach you.
Calif also described other possible routes an attacker could use to gain control of a WeChat account and propagate the worm, although these were attack routes it described rather than ones it tested. Importantly, the exploit itself doesn’t give the attacker control of the victim’s entire phone.
As per Tencent, the combined monthly active users of WeChat and Weixin stood at 1.439 billion as of June 30, 2026. It’s important to note that WeChat isn’t just a messaging and VoIP app, but also includes payments and mini programs inside it, which expands the potential risk of a worm-like attack such as WeWorm.
Calif reported the bug to Tencent in July, and Tencent subsequently released version 8.0.77 for Android and 8.0.76 for iOS on August 21. Calif said those releases mitigated its exploit, and on August 28, it confirmed that the exploit had also been blocked on Tencent’s servers.
Although Calif hasn’t published extensive technical details, it confirmed that it used AI to find a bug in WeChat’s VoIP stack and write the first remote code execution (RCE) exploit, which took about two days. After that, building the worm took another week.
Something like this hasn't always been this easy and quick to build. Calif's source says a worm at this scale previously required a larger team and months of work, whereas AI can now do most of the work.
Calif also said in its report that such advanced capabilities have already existed. However, AI has now put these capabilities in the hands of less-skilled threat actors, which were earlier only available to sophisticated and well-funded malicious parties.
Neither Tencent nor Calif has observed any exploitation of the bug in the wild, and no attacks have been reported. However, the whole demonstration goes to show how far AI has come in recent years.
And this isn't the first time we've seen AI unexpectedly find its way into real-world systems.
In a previous incident, an AI agent was simply asked to book a gym class. Instead, it exploited authorization flaws in the gym software provider’s GraphQL API and deleted another person’s reservation to move the user up the waitlist. The user never instructed the AI to hack, exploit, or bypass any security controls.
That incident is particularly worrying because it shows how little technical knowledge or malicious intent may be required. The user gave the AI a mundane task, and the agent independently found a way to exploit a vulnerability to achieve its goal.
Not long ago, OpenAI’s unreleased model escaped its testing environment and breached Hugging Face’s internal database to complete a challenge. In that case, the model identified and exploited vulnerabilities, moved laterally through the environment, escalated privileges, and eventually reached Hugging Face’s production database.
Threat actors have also been using AI to draft convincing phishing messages, making them increasingly difficult for the average user to spot. AI-generated emails can have better grammar, cleaner formatting, and more convincing language, removing many of the telltale signs that previously gave phishing emails away, such as typos, spelling mistakes, and awkward phrasing.
The common thread in all these incidents is that AI is lowering the barrier to sophisticated cyber activity. What once required considerable technical expertise and time can increasingly be accomplished by AI models in a matter of days, or even autonomously while trying to complete a seemingly harmless task.
With WeWorm’s demonstration, it’s now clear that AI could be used to find bugs and then exploit them with RCEs, potentially putting sophisticated cyber capabilities within reach of less-skilled threat actors.
The one bit of reassurance, though, is that, as Calif put it, “there are more good guys than bad guys, and if they are paying attention, AI gives the good guys the upper hand.” However, for everyday users, the reality is that AI is consistently making the digital threat landscape more dangerous.
An antivirus that scores 18/18 on AV-TEST for Windows and macOS, with top marks across all test categories
Passed every malware and drive-by download test we ran, quarantining threats automatically
Includes a junk cleaner, app uninstaller, and browser cleaner to keep your device running smoothly alongside the antivirus
/images/2023/10/11/best-data-removal-service.png)
/images/2023/06/01/incogni-review.png)
/images/2026/09/10/wechat-worm-ai-cyberattack.jpeg)
/images/2026/09/10/apple-watch-live-rewind-is-listening.png)
/images/2026/09/09/windows-age-verification.jpeg)
/images/2026/09/08/closeup_of_smartphone_with_plaid_banking_app_on_screen.jpg)
/images/2026/09/08/liquid-network-bitcoin-hack.png)
/images/2026/09/08/grindr-lawsuit-hiv-status.jpeg)
/authors/krishi_chowdhary_headshot.jpg){kind=small}
/authors/kalleigh-lane-new.jpg){kind=small}
/images/2022/09/22/logo-totalav.png){kind=logo}
/authors/krishi_chowdhary_headshot.jpg)