CVE-2026-93680

Langflow is a tool businesses use to build and run AI-powered workflows. This flaw is like a door that's supposed to need a key but can be opened without one — someone could get into the system without logging in. If exploited, an attacker could fully control the application and any data or workflows it manages.

This matters to you if you run Langflow OSS (also called Langflow), specifically any version from 1.0.0 through 1.12.8. The risk is highest if your Langflow instance is reachable from the internet rather than only accessible on an internal network.

There is no evidence yet that attackers are actively exploiting this flaw, and it is not on any government list of known-exploited vulnerabilities. However, the bypass is easy to trigger and requires no login of its own, so you should plan to update soon rather than treat it as optional — don't delay indefinitely, but this isn't an emergency fire drill either.

Confirm exposure by identifying all Langflow OSS deployments (pip installs and Docker containers) and checking their version against the 1.0.0–1.12.8 affected range. Upgrade to 1.13.0 or later: for pip deployments run pip install langflow --upgrade; for Docker deployments, pull the updated image referenced in the GitHub releases page (https://github.com/langflow-ai/langflow/releases) and redeploy. There is no documented breaking-change note from the vendor advisory, but test in staging before production rollout given the authentication-layer change. As a temporary workaround if patching is delayed, restrict network access to the Langflow management interface (e.g., firewall rules, VPN-only access, reverse proxy with authentication) to prevent unauthenticated reachability. No CISA KEV listing exists, so there is no federal remediation deadline. Monitor for unexpected admin actions, new user/API key creation, or configuration changes in Langflow logs as a sign of exploitation attempts.

CVSS 3.1 score is 9.8 (Critical), AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H — network-exploitable, no privileges or user interaction required, full impact on confidentiality, integrity, and availability. Weakness is CWE-287 (Improper Authentication), affecting versions 1.0.0–1.12.8, fixed in 1.13.0. EPSS predicts a 2.2% likelihood of exploitation in the next 30 days (82nd percentile), a relatively low absolute probability despite the high severity score. No public exploit code is known to exist, and the CVE is not listed in CISA's Known Exploited Vulnerabilities catalog. Press attention is newly emerging but should not be read as evidence of active exploitation.