CVE-2026-110193: Langflow: code injection in Agentic Assistant allows RCE

HIGH

Published October 29, 2026

CISO Take

IBM Langflow OSS 1.0.0 through 1.12.8 contains a code injection flaw (CWE-94) in which the Agentic Assistant can be driven to generate or submit component code containing unblocked patterns, giving any low-privileged authenticated user arbitrary OS command execution on the server. The CVSS score of 8.8 reflects network reachability, low complexity, and no user interaction. Langflow is a widely deployed low-code platform for building LLM agents and RAG workflows, and its servers commonly hold LLM API keys, database credentials, and vector store connections, so a compromise exposes the surrounding AI estate. No public exploit, Nuclei template, KEV listing, or EPSS score is available yet, so exploitation likelihood is unquantified, but the low privilege bar and the history of Langflow code-execution bugs make weaponization plausible. Upgrade to the fixed release named in the IBM advisory (https://www.ibm.com/support/pages/node/7294467), restrict Langflow to trusted users and networks in the meantime, and audit existing accounts and recent component code.

Sources:NVDATLASibm.com

What is the risk?

High. Exploitation needs only a low-privileged authenticated account, network access, and no user interaction, and it yields full confidentiality, integrity, and availability impact. The risk is highest for multi-user or internet-exposed Langflow instances and for deployments that grant broad self-registration or shared accounts. No public exploit, Nuclei template, KEV listing, or EPSS data was available at publication, so likelihood of exploitation is currently unquantified. Confidence in the fix scope is HIGH, based on the vendor advisory; confidence in the exploit mechanics is MEDIUM, since only the CVE description is available.

How does the attack unfold?

Initial Access

Attacker logs in to Langflow with a valid low-privileged account.

AML.T0012

Code Injection

Attacker has the Agentic Assistant generate or submit component code using patterns the validator does not block.

AML.T0053

Execution

The injected code runs on the server and executes arbitrary OS commands.

AML.T0050

Impact

Attacker steals API keys and credentials, tampers with flows, and pivots into the internal network.

AML.T0049

Initial Access

Attacker logs in to Langflow with a valid low-privileged account.

AML.T0012

Code Injection

Attacker has the Agentic Assistant generate or submit component code using patterns the validator does not block.

AML.T0053

Execution

The injected code runs on the server and executes arbitrary OS commands.

AML.T0050

Impact

Attacker steals API keys and credentials, tampers with flows, and pivots into the internal network.

AML.T0049

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
Langflow	pip	—	No patch
156.0KPushed 6d ago18% patched~66d to patchFull package profile →

Do you use Langflow? You're affected.

How severe is it?

CVSS 3.1

8.8 / 10

EPSS

0.8%

chance of exploitation in 30 days

Higher than 56% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

No known exploitation

Sophistication

Moderate

What is the attack surface?

AVNetwork

ACLow

PRLow

UINone

SUnchanged

CHigh

IHigh

AHigh

What should I do?

1 step

1) Upgrade to the patched version listed in IBM advisory 7294467. 2) Until patched, disable or restrict access to the Agentic Assistant feature, and limit Langflow accounts to trusted users. 3) Place Langflow behind a VPN or an authenticating reverse proxy and do not expose it directly to the internet. 4) Run Langflow in a hardened container as a non-root user, with restricted egress and a minimal filesystem. 5) Rotate LLM API keys and connector credentials stored in Langflow if compromise is suspected. 6) Detect by monitoring for unexpected child processes spawned by the Langflow process (shells, curl, wget), unusual outbound connections, and suspicious custom component code or Agentic Assistant requests in application logs.

How is it classified?

Code ExecutionPrompt InjectionFrameworkAgentAML.T0012- Valid Accounts AML.T0049- Exploit Public-Facing Application AML.T0050- Command and Scripting Interpreter AML.T0053- AI Agent Tool Invocation AML.T0102- Generate Malicious Commands

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act

Art. 15-Accuracy, robustness and cybersecurity

ISO 42001

A.6.2.6-AI system operation and monitoring

NIST AI RMF

MANAGE 2.3-Responses to previously unknown risks are managed

OWASP LLM Top 10

LLM05-Improper Output HandlingLLM06-Excessive Agency

Frequently Asked Questions

What is CVE-2026-110193?

Is CVE-2026-110193 actively exploited?

No confirmed active exploitation of CVE-2026-110193 has been reported, but organizations should still patch proactively.

How to fix CVE-2026-110193?

1) Upgrade to the patched version listed in IBM advisory 7294467. 2) Until patched, disable or restrict access to the Agentic Assistant feature, and limit Langflow accounts to trusted users. 3) Place Langflow behind a VPN or an authenticating reverse proxy and do not expose it directly to the internet. 4) Run Langflow in a hardened container as a non-root user, with restricted egress and a minimal filesystem. 5) Rotate LLM API keys and connector credentials stored in Langflow if compromise is suspected. 6) Detect by monitoring for unexpected child processes spawned by the Langflow process (shells, curl, wget), unusual outbound connections, and suspicious custom component code or Agentic Assistant requests in application logs.

What systems are affected by CVE-2026-110193?

This vulnerability affects the following AI/ML architecture patterns: agent frameworks, low-code LLM workflow builders, RAG pipelines, LLM orchestration servers.

What is the CVSS score for CVE-2026-110193?

CVE-2026-110193 has a CVSS v3.1 base score of 8.8 (HIGH). The EPSS exploitation probability is 0.78%.

What is the AI security impact?

Affected AI Architectures

agent frameworkslow-code LLM workflow buildersRAG pipelinesLLM orchestration servers

MITRE ATLAS Techniques

AML.T0012Valid Accounts

AML.T0049Exploit Public-Facing Application

AML.T0050Command and Scripting Interpreter

AML.T0053AI Agent Tool Invocation

AML.T0102Generate Malicious Commands

Compliance Controls Affected

EU AI Act: Art. 15

ISO 42001: A.6.2.6

NIST AI RMF: MANAGE 2.3

OWASP LLM Top 10: LLM05, LLM06

What are the technical details?

Original Advisory

IBM Langflow OSS 1.0.0 through 1.12.8 could allow a remote attacker with a low-privileged authenticated account to cause the Agentic Assistant to generate or submit component code using these unblocked patterns, resulting in arbitrary operating system command execution on the server.

Exploitation Scenario

An attacker obtains or registers a low-privileged account on an exposed Langflow instance, for example through credential stuffing or an open signup. They prompt the Agentic Assistant to create or submit a custom component whose code uses a pattern the validator does not block, such as a dynamic import or an indirect call to os or subprocess. When the component is built or executed, the code runs on the server and the attacker gets a reverse shell. They then read environment variables and stored credentials for LLM providers and data stores, and move laterally into internal data sources.

Weaknesses (CWE)

CWE-94Improper Control of Generation of Code ('Code Injection')Primary CWE-94Improper Control of Generation of Code ('Code Injection')

CWE-94 — Improper Control of Generation of Code ('Code Injection'): 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.