CVE-2026-90535

Flowise includes a text-to-speech feature. This bug lets an attacker send a “stop/abort” request that crashes or halts text-to-speech activity for other users, kind of like someone having the ability to hit “stop” on a coworker’s voice reading in the middle of it.

The result is not data theft—it’s interruption of service, which can break customer sessions or internal workflows.

This matters if you run Flowise (a text-to-speech/chat workflow platform) and your server is reachable over the network. It’s a risk even without login.

It’s specifically a concern when Flowise is running a version earlier than 3.1.4 and the relevant API endpoint is exposed, especially from the public internet.

This is AMBER because an attacker can trigger a denial-of-service condition without authentication and without user interaction, as long as the Flowise endpoint is exposed. While there’s no confirmed public exploit code and no KEV listing, the default-config reachability and simple “stop” mechanism make it a practical disruption risk.

Act quickly by upgrading to 3.1.4 and limiting external access to the API if you can’t upgrade immediately.

1. Confirm exposure

   • Identify your Flowise deployment version (look at the installed version in your setup/docs/UI, or in the container/image tag you run).
   • Determine whether Flowise is reachable from the internet.
   • Confirm whether the API route is exposed externally (for example, any public gateway/reverse proxy that forwards /api/v1/… to the internet).

2. Confirm likely affected state

   • Treat the system as affected if: Flowise version is earlier than 3.1.4 AND the /api/v1/text-to-speech/abort endpoint can be reached over the network.

3. Upgrade to the fixed version

   • Upgrade Flowise to 3.1.4 (or later). This is the vendor-fixed version for CVE-2026-90535.

4. If patching is delayed (temporary workaround)

   • Lock down network access to Flowise’s API so outsiders can’t reach it (e.g., restrict inbound traffic to your admin IPs/VPN, or only allow access from trusted internal networks).
   • If you run a reverse proxy or firewall, ensure it does not publish the abort endpoint to the public internet.

5. What to monitor after the change

   • Check application logs for repeated calls to the text-to-speech abort endpoint.
   • Monitor for sudden increases in text-to-speech interruptions or aborted sessions.
   • Ensure access control changes prevent unauthenticated requests from reaching the endpoint.

6. Timing

   • No KEV due date was provided in the findings.

Severity and weakness: This is a denial-of-service issue (CWE-862) in Flowise’s text-to-speech abort handling. The weakness allows unauthenticated network requests to terminate/abort active text-to-speech operations.

Mechanism / attack vector: The /api/v1/text-to-speech/abort endpoint accepts user-supplied chatflowId and chatId and does not verify ownership before allowing an abort. As a result, an attacker can send a crafted request to stop text-to-speech activity for any user.

Exposure conditions: No authentication and no user interaction are required. The findings indicate the endpoint is reachable in default configuration, meaning exposure is primarily determined by whether the API endpoint is reachable from the network.

Exploitation status: No public exploit code is recorded in the findings, and CISA KEV does not list this vulnerability. However, the attack is straightforward (send an abort request), and the availability impact is direct.

Fix: Vendor remediation indicates Flowise is fixed in version 3.1.4.

Predicted likelihood: EPSS was provided as a prediction in the findings, but this is not confirmed exploitation activity.