WeWorm: How AI Built a Zero-Click Worm That Hacks WeChat Without Answering a Call

A security research team has demonstrated a WeChat attack that can take over an account while an incoming call is still ringing. The researchers say AI helped them find the flaw and build the exploit in days.

Researchers demonstrated WeWorm moving between Android and iOS devices through WeChat calls. Image: FactsWings illustration.

The attack starts with something ordinary

WeWorm does not begin with the kind of suspicious message many people have learned to avoid.

There is no fake prize. No strange attachment. No request for a password.

The demonstration begins with a WeChat call.

Calif says its researchers were able to exploit a vulnerability while the incoming call was still ringing. The person receiving the call did not need to answer it or interact with the device.

That is what makes the vulnerability a zero-click problem.

In ordinary cyberattacks, users are often part of the security boundary. They can refuse a suspicious download or close a suspicious message.

A zero-click vulnerability can remove that choice.

What happened to the account?

According to Calif, a successful exploitation gave the researchers control of the targeted WeChat account.

They say they could read and send messages, make calls and act on behalf of the compromised account.

The researchers then used the compromised account to call another phone.

That second phone could be compromised in the same way, turning the victim into the next potential attacker.

Calif demonstrated the chain using three phones: an Android device called the attacker, an iPhone as the first victim, and another Android phone as the next target.

The demonstration showed that the attack could cross between the two major mobile operating systems.

Why a messaging app makes the problem worse

A compromised messaging account has something that many other compromised devices do not have: a ready-made network of contacts.

People generally trust calls and messages from people they already know.

If one account is compromised, that trust can become part of the attack.

Calif says its demonstration required the attacker to be on the victim’s friend list. But the researchers argue that this is not necessarily a strong barrier because an attacker could first compromise another account and use that trusted account to reach additional contacts.

This is the basic idea behind the worm.

The bug was inside WeChat’s calling system

Calif describes the underlying flaw as a memory-corruption issue in WeChat’s VoIP stack.

VoIP is the technology used to carry voice calls over an internet connection.

The researchers have deliberately withheld the technical details of the vulnerability.

They say they plan to present the full analysis at a future security conference.

That means there is an important limit to what can currently be said publicly. The existence and impact of the demonstrated vulnerability have been described, but the complete exploit chain has not been released.

Then there is the AI question

The most surprising part of the research may be the time it took.

Calif says its team worked with AI to find the vulnerability and write its first remote-code-execution exploit in about two days.

Building the worm demonstration took another week.

That is much faster than the kind of long development process normally associated with sophisticated vulnerability research.

The researchers are careful to say that humans were still involved. Their team decided what to target and how to test the work safely.

But AI handled enough of the work to change the timetable.

That is the bigger story.

AI could make both sides faster

The same technology that helps researchers find vulnerabilities can also potentially help criminals.

Security teams can use AI to inspect code, analyse large amounts of data, identify suspicious behaviour and search for weaknesses.

Attackers can try to automate similar tasks.

The result is a new race: can defenders find and fix a vulnerability before someone else turns it into an attack?

Calif argues that AI could ultimately help defenders because it can also make vulnerability discovery and fixing faster.

But the WeWorm demonstration shows why that race is becoming more urgent.

Tencent was warned before the research went public

Calif says it reported the WeChat vulnerability to Tencent in July.

The researchers’ published timeline gives a more detailed picture of what happened next.

Is WeChat already infected?

There is no evidence in the published research of a global WeWorm outbreak.

This is a demonstration of what the vulnerability could do, not evidence that the worm was released against the public.

That distinction matters.

The researchers say Tencent has mitigated their exploit for all users. Independent reporting also says there was no known active exploitation of the vulnerability when the story was published.

So users should not read the existence of WeWorm as proof that their phones have been infected.

What users should do now

There is no complicated security checklist for ordinary WeChat users.

The most important step is to keep WeChat updated.

The same applies to your phone’s operating system.

Do not install unofficial or modified versions of messaging applications, and be cautious if a trusted account suddenly behaves in an unusual way.

But there is one important caveat: users cannot always protect themselves from a zero-click vulnerability through careful clicking habits. If the vulnerable software processes malicious data automatically, the software developer has to fix the underlying problem.

This is bigger than one app

WeWorm is a warning about the hidden attack surface inside modern messaging applications.

A messaging app may look simple on the screen, but behind it are systems for calls, video, media, notifications, file processing and network communication.

Every one of those systems can contain vulnerabilities.

Calif says its WeChat research is part of a wider effort to investigate zero-click attack surfaces in mobile messaging applications.

That could make similar research increasingly important for other communication platforms.

The uncomfortable lesson from WeWorm

The most worrying part of WeWorm is not simply that a serious bug existed.

Software vulnerabilities have always existed.

The worrying part is the speed.

Calif says AI helped its researchers move from finding the vulnerability to a working RCE exploit in roughly two days.

If AI continues to improve at this pace, the time between a vulnerability being discovered and someone being able to exploit it could become much shorter.

That puts pressure on developers to find vulnerabilities earlier, patch them faster and make automatic security updates more reliable.

It also gives defenders a reason to use AI themselves.

The future of cybersecurity may not be a simple battle between humans and machines. It may be a race between AI-assisted attackers and AI-assisted defenders.

What is confirmed — and what is not

Confirmed: Calif demonstrated a zero-click WeChat worm across Android and iOS. The company says the vulnerability is a memory-corruption issue in WeChat’s VoIP stack. The researchers say AI helped them discover the bug and build the first RCE exploit in about two days.

Confirmed: Calif reported the issue to Tencent, and the company says WeChat updates and server-side mitigation were deployed.

Not established: There is no evidence from the published research that WeWorm caused a worldwide outbreak or that millions of ordinary users were actually infected.

Not public: The full technical exploit details have not been released by the researchers.

Keeping those facts separate is important. A dramatic cybersecurity story can be useful only when the confirmed facts are not mixed with speculation.

Sources and evidence

FactsWings checked the core claims in this report against the original research disclosure and independent security reporting.

Frequently asked questions