Impact

There is an unauthenticated remote code execution vulnerability in React Server Components.

We recommend upgrading immediately.

The vulnerability is present in versions 19.0.0, 19.1.0, 19.1.1, and 19.2.0 of: react-server-dom-webpack react-server-dom-parcel react-server-dom-turbopack

Patches

A fix was introduced in versions 19.0.1, 19.1.2, and 19.2.1. If you are using any of the above packages please upgrade to any of the fixed versions immediately.

If your app’s React code does not use a server, your app is not affected by this vulnerability. If your app does not use a framework, bundler, or bundler plugin that supports React Server Components, your app is not affected by this vulnerability.

References

See the blog post for more information and upgrade instructions.

In PyTorch <=2.4.1, the RemoteModule has Deserialization RCE. NOTE: this is disputed by multiple parties because this is intended behavior in PyTorch distributed computing.

An OS command injection vulnerability in MetaGPT 0.8.1 allows an attacker to execute arbitrary commands via the path argument of RepoParser.rebuildclassviews() in metagpt/repoparser.py.

Open GenAI Stack (aka ogx-ai) 2026-06-11, as used in the Meta AI backend for WhatsApp and other products, allows code execution because prompt injection (with Jinja2 template syntax) can be used to achieve server-side expression evaluation without sanitization.

A use after free in hermes, while emitting certain error messages, prior to commit d86e185e485b6330216dee8e854455c694e3a36e allows attackers to potentially execute arbitrary code via crafted JavaScript. Note that this is only exploitable if the application using Hermes permits evaluation of untrusted JavaScript. Hence, most React Native applications are not affected.

Prior to v74.0.0.878.1682 of Meta Horizon OS, MediaSyncJobReceiver could be induced to send a privileged PendingIntent including a com.oculus.vrshell CallerIdentity to an arbitrary application listening via NotificationListenerService. That would allow the application to impersonate the com.oculus.vrshell package, as well as packages signed with the same key, towards any endpoint within the OS that uses CallerIdentity authentication.

Prior to v66.0.0.733.524 of Meta Horizon OS, OVRMediaService could be induced to send a privileged PendingIntent including a com.oculus.horizon CallerIdentity to an arbitrary application registering for com.oculus.systemactivities.SCREENSHOT via a broadcast receiver. That would allow the application to impersonate the com.oculus.horizon package towards any endpoint within the OS that uses CallerIdentity authentication.

Impact

It was found that the fix to address CVE-2025-55184 in React Server Components was incomplete and does not prevent a denial of service attack in a specific case.

We recommend updating immediately.

The vulnerability exists in versions 19.0.2, 19.1.3, and 19.2.2 of:

- react-server-dom-webpack - react-server-dom-parcel - react-server-dom-turbopack

These issues are present in the patches published on December 11th, 2025.

Patches

Fixes were back ported to versions 19.0.3, 19.1.4, and 19.2.3.

If you are using any of the above packages please upgrade to any of the fixed versions immediately.

If your app’s React code does not use a server, your app is not affected by this vulnerability. If your app does not use a framework, bundler, or bundler plugin that supports React Server Components, your app is not affected by this vulnerability.

References

See the blog post for more information and upgrade instructions.

A path traversal vulnerability in the SPO extension of MetaGPT 0.8.1 allows an attacker to read arbitrary files via the FILENAME value used by setfilename() and loadmetadata() in metagpt/ext/spo/utils/load.py. The vulnerable code joins the attacker-controlled FILENAME value with the settings directory and opens the resulting path without validating that the resolved path remains within the intended directory.

In proxygen from v2024.10.28.00 until v2026.09.28.00, WebTransportImpl::terminateSessionStreams (WebTransportImpl::destroy in releases before v2025.08.18.00) failed to unregister read callbacks for streams that were no longer open before destroying them. The transport could then invoke a read callback that had been freed.

In proxygen from v2026.04.06.00 until v2026.09.28.00, QuicWtSession::closeSession accesses its member fields after calling the base QuicWtSessionBase::closeSession method. The base method notifies the session handler, which may release the last reference to the session and destroy it.

The Meta Display Block plugin for WordPress is vulnerable to Stored Cross-Site Scripting via the Meta Display Block in all versions up to, and including, 1.0.0 due to insufficient input sanitization and output escaping. This makes it possible for authenticated attackers, with Contributor-level access and above, to inject arbitrary web scripts in pages that will execute whenever a user accesses an injected page.

Incomplete authorization of linked device synchronization messages in WhatsApp for iOS prior to v2.25.21.73, WhatsApp Business for iOS v2.25.21.78, and WhatsApp for Mac v2.25.21.78 could have allowed an unrelated user to trigger processing of content from an arbitrary URL on a target’s device. We assess that this vulnerability, in combination with an OS-level vulnerability on Apple platforms (CVE-2025-43300), may have been exploited in a sophisticated attack against specific targeted users.

In proxygen from v2024.10.28.00 until v2026.09.28.00, the HTTPTransaction::onWebTransportUniStream and HTTPTransaction::onWebTransportBidiStream APIs could return stream handles that the stream handler had already freed. HQSession then installed those handles as transport read callbacks, which could lead to use of freed memory.

A vulnerability, which was classified as problematic, has been found in PyTorch 2.6.0+cu124. Affected by this issue is the function torch.mkldnnmaxpool2d. The manipulation leads to denial of service. An attack has to be approached locally. The exploit has been disclosed to the public and may be used.

A vulnerability was found in PyTorch 2.6.0+cu124. It has been rated as problematic. Affected by this issue is the function torch.cuda.nccl.reduce of the file torch/cuda/nccl.py. The manipulation leads to denial of service. It is possible to launch the attack on the local host. The exploit has been disclosed to the public and may be used. The patch is identified as 5827d2061dcb4acd05ac5f8e65d8693a481ba0f5. It is recommended to apply a patch to fix this issue.

Meta recently patched a high-severity vulnerability in WhatsApp for Windows (CVE-2025-30401) that allowed attackers to execute malicious code on a target’s machine by sending spoofed file types. The exploit didn’t require the user to do anything beyond opening what looked like a harmless file. This makes it particularly dangerous for less tech-savvy users and internal teams not trained in threat detection. What is known: \- Affected all WhatsApp for Windows versions prior to 2.2450.6 \- Exploited using specially crafted files with manipulated extensions \- Meta confirmed active exploitation before the patch was issued It’s a reminder that even popular desktop apps can become threat vectors when spoofing and user trust intersect. Were you or someone you know affected by this? Or have you seen this exploited in the wild in an org you work with?