CVE-2026-18729: Langflow: authenticated RCE via unsafe code generation

HIGHPoC AVAILABLE

Published August 28, 2026

CISO Take

IBM Langflow OSS versions 1.0.0 through 1.11.1 contain a code injection flaw (CWE-94) that lets any authenticated user — even one with low privileges — execute arbitrary code on the server, scoring 8.8 (High) with full compromise of confidentiality, integrity, and availability. There's no EPSS score, no public exploit, no Nuclei template, and it's not in CISA KEV yet, so this isn't a drop-everything breaking alert, but Langflow sits at the center of many low-code agent and RAG pipelines and a low-privilege account is a trivial bar for insider threats, leaked credentials, or open self-service signups. Because Langflow servers routinely hold API keys for LLM providers, vector databases, and internal tools, a single compromised low-priv account can escalate to full server takeover and lateral movement into connected AI infrastructure. Patch to a fixed release above 1.11.1 immediately, and until then restrict Langflow access to trusted internal users only, disable open self-registration, and monitor for anomalous custom-component or code-execution activity in Langflow logs.

Sources:NVDibm.comATLAS

What is the risk?

High severity (CVSS 8.8) with low attack complexity and no user interaction, but it requires low-privilege authentication, which meaningfully narrows the exploitable population compared to fully unauthenticated RCE. No EPSS data, no known public exploit code, no scanner template, and absence from CISA KEV all suggest exploitation is not yet observed in the wild. However, the impact ceiling is total (C:H/I:H/A:H) and Langflow is a widely deployed low-code AI orchestration tool, so any environment with loose account provisioning, shared credentials, or self-service signup meaningfully raises real-world exploitability. Treat this as a high-priority patch item, not an emergency, unless your Langflow instance allows broad or self-service account creation.

How does the attack unfold?

Initial Access

Attacker obtains a low-privileged authenticated Langflow account via self-registration, leaked credentials, or an over-permissive invite.

AML.T0012

Exploitation

Attacker crafts a malicious custom component that abuses Langflow's improperly controlled code-generation feature to inject arbitrary Python code.

AML.T0049

Execution

The injected code runs on the Langflow server with the service account's privileges, granting the attacker command execution.

AML.T0050

Impact

Attacker harvests connected LLM/vector-store credentials, tampers with other users' AI workflows, and pivots to reachable infrastructure.

AML.T0025

Initial Access

Attacker obtains a low-privileged authenticated Langflow account via self-registration, leaked credentials, or an over-permissive invite.

AML.T0012

Exploitation

Attacker crafts a malicious custom component that abuses Langflow's improperly controlled code-generation feature to inject arbitrary Python code.

AML.T0049

Execution

The injected code runs on the Langflow server with the service account's privileges, granting the attacker command execution.

AML.T0050

Impact

Attacker harvests connected LLM/vector-store credentials, tampers with other users' AI workflows, and pivots to reachable infrastructure.

AML.T0025

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
Langflow	pip	—	No patch
156.2KPushed 3d ago18% patched~75d to patchFull package profile →

Do you use Langflow? You're affected.

How severe is it?

CVSS 3.1

8.8 / 10

EPSS

1.6%

chance of exploitation in 30 days

Higher than 75% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

Exploit Available

Exploitation: MEDIUM

Sophistication

Moderate

Exploitation Confidence

medium

✓Weaponized Metasploit module (exploit/multi/http/langflow_auth_rce_cve_2026_18729)

Composite signal derived from CISA KEV, VulnCheck KEV, CISA SSVC, EPSS, Metasploit, Exploit-DB, trickest/cve, Nuclei templates, and inthewild.io exploitation reports.

What is the attack surface?

AVNetwork

ACLow

PRLow

UINone

SUnchanged

CHigh

IHigh

AHigh

What should I do?

1 step

Upgrade Langflow OSS to a version beyond 1.11.1 that fixes the improper code-generation control (check IBM's advisory for the exact patched release). Until patched, restrict Langflow access to a small set of trusted, vetted users and disable open or self-service account registration to raise the bar for obtaining the low-privilege access this exploit requires. Audit existing user accounts and roles for unnecessary access, rotate any secrets/API keys stored in or accessible from the Langflow instance, and run Langflow under a least-privilege service account with restricted filesystem/network egress to limit blast radius if exploited. For detection, monitor Langflow server logs and EDR telemetry for unexpected child processes, outbound connections, or file writes originating from the Langflow process, particularly following custom-component creation or edits.

What does CISA's SSVC say?

DecisionTrack

Exploitationnone

AutomatableNo

Technical Impacttotal

Source: CISA Vulnrichment (SSVC v2.0). Decision based on the CISA Coordinator decision tree.

How is it classified?

Code ExecutionAuth BypassFrameworkAgentAML.T0012- Valid Accounts AML.T0049- Exploit Public-Facing Application AML.T0050- Command and Scripting Interpreter

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act

Article 15-Accuracy, robustness and cybersecurity

ISO 42001

A.6.2.6-AI system security

NIST AI RMF

MANAGE-4.1-Risks and benefits are monitored and mitigations documented

OWASP LLM Top 10

LLM07-Insecure Plugin/Component Design

Frequently Asked Questions

What is CVE-2026-18729?

Is CVE-2026-18729 actively exploited?

A weaponized Metasploit module (exploit/multi/http/langflow_auth_rce_cve_2026_18729) exists for CVE-2026-18729, meaning the exploit is point-and-click and the risk of opportunistic exploitation is high.

How to fix CVE-2026-18729?

Upgrade Langflow OSS to a version beyond 1.11.1 that fixes the improper code-generation control (check IBM's advisory for the exact patched release). Until patched, restrict Langflow access to a small set of trusted, vetted users and disable open or self-service account registration to raise the bar for obtaining the low-privilege access this exploit requires. Audit existing user accounts and roles for unnecessary access, rotate any secrets/API keys stored in or accessible from the Langflow instance, and run Langflow under a least-privilege service account with restricted filesystem/network egress to limit blast radius if exploited. For detection, monitor Langflow server logs and EDR telemetry for unexpected child processes, outbound connections, or file writes originating from the Langflow process, particularly following custom-component creation or edits.

What systems are affected by CVE-2026-18729?

This vulnerability affects the following AI/ML architecture patterns: agent frameworks, RAG pipelines.

What is the CVSS score for CVE-2026-18729?

CVE-2026-18729 has a CVSS v3.1 base score of 8.8 (HIGH). The EPSS exploitation probability is 1.55%.

What is the AI security impact?

Affected AI Architectures

agent frameworksRAG pipelines

MITRE ATLAS Techniques

AML.T0012Valid Accounts

AML.T0049Exploit Public-Facing Application

AML.T0050Command and Scripting Interpreter

Compliance Controls Affected

EU AI Act: Article 15

ISO 42001: A.6.2.6

NIST AI RMF: MANAGE-4.1

OWASP LLM Top 10: LLM07

What are the technical details?

Original Advisory

IBM Langflow OSS 1.0.0 through 1.11.1 could allow a remote authenticated attacker to execute arbitrary code due to improper control of generation of code.

Exploitation Scenario

An attacker obtains a low-privilege Langflow account — through self-service signup, a phished/leaked credential, or an over-permissioned internal invite — and logs into the platform. Using Langflow's custom component or code-generation feature, they craft a component that embeds malicious Python payloads instead of legitimate transformation logic. Because Langflow fails to properly control what generated code can do, the payload executes on the server with the Langflow service's own privileges rather than being contained. From there, the attacker harvests LLM provider API keys and vector database credentials stored in the environment, tampers with other users' agent workflows, and pivots to other systems reachable from the compromised host — turning a single low-privilege account into full platform and pipeline compromise.

Weaknesses (CWE)

CWE-94Improper Control of Generation of CodePrimary CWE-94Improper Control of Generation of Code

CWE-94 — Improper Control of Generation of Code: The product constructs all or part of a code segment using externally-influenced input from an upstream component, but it does not neutralize or incorrectly neutralizes special elements that could modify the syntax or behavior of the intended code segment.

[Architecture and Design] Refactor your program so that you do not have to dynamically generate code.
[Architecture and Design] Run your code in a "jail" or similar sandbox environment that enforces strict boundaries between the process and the operating system. This may effectively restrict which code can be executed by your product. Examples include the Unix chroot jail and AppArmor. In general, managed code may provide some protection. This may not be a feasible solution, and it only limits the impact to the operating system; the rest of your application may still be subject to compromise. Be careful to avoid CWE-243 and other weaknesses related to jails.

Source: MITRE CWE corpus.