CVE-2026-4502: Langflow path traversal enables

CISO Take

IBM Langflow Desktop versions 1.2.0 through 1.8.4 contain a path traversal flaw (CWE-22) that lets any authenticated user write arbitrary files anywhere on the host by embedding dot-dot sequences in URL requests — no admin rights needed, no user interaction required. While the raw EPSS score is low at 0.00052, it sits in the top 84th percentile for exploitation likelihood, and the HIGH integrity impact (CVSS I:H) combined with a network-accessible, low-complexity attack vector makes this a practical weapon in any environment where Langflow is reachable over a network. In LLM orchestration contexts, arbitrary file write is rarely just file write — an attacker who overwrites a Langflow flow definition, agent tool script, or startup file converts this into de facto code execution at the next pipeline run. No public exploit exists and SSVC is TRACK, but the low bar for exploitation means teams should prioritize patching and immediately verify that Langflow instances are not directly network-exposed with minimal access controls.

Sources: NVD EPSS CISA KEV ATLAS

What is the risk?

CVSS 6.5 Medium understates operational risk in AI/ML environments. Arbitrary file write in an LLM orchestration framework can escalate to code execution if an attacker targets Python scripts, flow configs, or environment files loaded at runtime — a well-known post-exploitation pattern in web application path traversal attacks. Low attack complexity and low privilege requirement make this reachable by any authenticated user, not just administrators, widening the threat actor pool to include insider threats and compromised service accounts. No KEV listing and no public exploit anchor this at TRACK, but the EPSS 84th percentile signals meaningful real-world exploitation interest.

How does the attack unfold?

Initial Access

Attacker authenticates to Langflow Desktop with a low-privilege account obtained via credential theft, default credentials, or compromised service account.

AML.T0012

Exploitation

Attacker sends a specially crafted URL request containing dot-dot sequences (/../) to a Langflow file-handling endpoint, bypassing path sanitization to target arbitrary filesystem locations.

AML.T0049

Persistence via Config Poisoning

Malicious content is written to Langflow flow definitions or agent tool configuration files, injecting adversarial logic or code that executes automatically on the next pipeline invocation.

AML.T0081

Impact

Modified configurations silently exfiltrate LLM API keys and conversation data, redirect model queries to attacker-controlled endpoints, or achieve code execution within the Langflow process context.

AML.T0048

Initial Access

Attacker authenticates to Langflow Desktop with a low-privilege account obtained via credential theft, default credentials, or compromised service account.

AML.T0012

Exploitation

Attacker sends a specially crafted URL request containing dot-dot sequences (/../) to a Langflow file-handling endpoint, bypassing path sanitization to target arbitrary filesystem locations.

AML.T0049

Persistence via Config Poisoning

Malicious content is written to Langflow flow definitions or agent tool configuration files, injecting adversarial logic or code that executes automatically on the next pipeline invocation.

AML.T0081

Impact

Modified configurations silently exfiltrate LLM API keys and conversation data, redirect model queries to attacker-controlled endpoints, or achieve code execution within the Langflow process context.

AML.T0048

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
Langflow	pip	—	No patch
149.9K Pushed 3d ago 40% patched ~67d to patch Full package profile →

Do you use Langflow? You're affected.

How severe is it?

CVSS 3.1

6.5 / 10

EPSS

0.3%

chance of exploitation in 30 days

Higher than 19% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

No known exploitation

Sophistication

Trivial

What is the attack surface?

AV Network

AC Low

PR Low

UI None

S Unchanged

C None

I High

A None

What should I do?

5 steps

Patch: Upgrade IBM Langflow Desktop beyond version 1.8.4; review IBM's advisory at ibm.com/support/pages/node/7271097 for the patched release.
Network isolation: Ensure no Langflow instance is directly internet-facing; place behind authenticated proxies or VPN even for internal deployments.
Least privilege: Run Langflow processes under a dedicated OS account with minimal file system permissions to bound the blast radius of a successful traversal.
File integrity monitoring: Deploy FIM rules on Langflow config directories, flow definition paths, and any Python files loaded at pipeline startup — alert on unexpected writes.
Audit authenticated users: Review which accounts have Langflow access and remove stale or overly broad access immediately.

What does CISA's SSVC say?

Decision Track

Exploitation none

Automatable No

Technical Impact partial

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

How is it classified?

Code Execution Supply Chain Framework Agent AML.T0049 - Exploit Public-Facing Application AML.T0055 - Unsecured Credentials AML.T0081 - Modify AI Agent Configuration AML.T0110 - AI Agent Tool Poisoning

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act

Article 9 - Risk Management System

ISO 42001

A.6.2.5 - Security of AI System Components

NIST AI RMF

MANAGE 2.4 - Risks and incidents are managed according to organizational policies

OWASP LLM Top 10

LLM03 - Supply Chain Vulnerabilities

Frequently Asked Questions

What is CVE-2026-4502?

IBM Langflow Desktop versions 1.2.0 through 1.8.4 contain a path traversal flaw (CWE-22) that lets any authenticated user write arbitrary files anywhere on the host by embedding dot-dot sequences in URL requests — no admin rights needed, no user interaction required. While the raw EPSS score is low at 0.00052, it sits in the top 84th percentile for exploitation likelihood, and the HIGH integrity impact (CVSS I:H) combined with a network-accessible, low-complexity attack vector makes this a practical weapon in any environment where Langflow is reachable over a network. In LLM orchestration contexts, arbitrary file write is rarely just file write — an attacker who overwrites a Langflow flow definition, agent tool script, or startup file converts this into de facto code execution at the next pipeline run. No public exploit exists and SSVC is TRACK, but the low bar for exploitation means teams should prioritize patching and immediately verify that Langflow instances are not directly network-exposed with minimal access controls.

Is CVE-2026-4502 actively exploited?

No confirmed active exploitation of CVE-2026-4502 has been reported, but organizations should still patch proactively.

How to fix CVE-2026-4502?

1. Patch: Upgrade IBM Langflow Desktop beyond version 1.8.4; review IBM's advisory at ibm.com/support/pages/node/7271097 for the patched release. 2. Network isolation: Ensure no Langflow instance is directly internet-facing; place behind authenticated proxies or VPN even for internal deployments. 3. Least privilege: Run Langflow processes under a dedicated OS account with minimal file system permissions to bound the blast radius of a successful traversal. 4. File integrity monitoring: Deploy FIM rules on Langflow config directories, flow definition paths, and any Python files loaded at pipeline startup — alert on unexpected writes. 5. Audit authenticated users: Review which accounts have Langflow access and remove stale or overly broad access immediately.

What systems are affected by CVE-2026-4502?

This vulnerability affects the following AI/ML architecture patterns: agent frameworks, LLM orchestration pipelines, multi-agent systems.

What is the CVSS score for CVE-2026-4502?

CVE-2026-4502 has a CVSS v3.1 base score of 6.5 (MEDIUM). The EPSS exploitation probability is 0.27%.

What is the AI security impact?

Affected AI Architectures

agent frameworksLLM orchestration pipelinesmulti-agent systems

MITRE ATLAS Techniques

AML.T0049 Exploit Public-Facing Application

AML.T0055 Unsecured Credentials

AML.T0081 Modify AI Agent Configuration

AML.T0110 AI Agent Tool Poisoning

Compliance Controls Affected

EU AI Act: Article 9

ISO 42001: A.6.2.5

NIST AI RMF: MANAGE 2.4

OWASP LLM Top 10: LLM03

What are the technical details?

Original Advisory

IBM Langflow Desktop 1.2.0 through 1.8.4 Langflow could allow an authenticated attacker to traverse directories on the system. An attacker could send a specially crafted URL request containing "dot dot" sequences (/../) to write arbitrary files on the system.

Exploitation Scenario

An attacker with a compromised low-privilege Langflow account (obtained via credential stuffing or insider access) crafts a POST request to a Langflow file-handling endpoint embedding a traversal payload such as /../../../opt/langflow/flows/main_agent.json. The server fails to normalize the path before writing, overwriting the primary agent flow definition with attacker-controlled JSON that adds a new tool call routing all LLM responses to an external webhook. On the next legitimate user interaction with the Langflow agent, the modified flow silently exfiltrates conversation content and any API keys in the agent's environment to the adversary's endpoint — all without triggering visible errors in the Langflow UI.

Weaknesses (CWE)

CWE-22 Improper Limitation of a Pathname to a Restricted Directory ('Path Traversal') Primary CWE-22 Improper Limitation of a Pathname to a Restricted Directory ('Path Traversal')

CWE-22 — Improper Limitation of a Pathname to a Restricted Directory ('Path Traversal'): The product uses external input to construct a pathname that is intended to identify a file or directory that is located underneath a restricted parent directory, but the product does not properly neutralize special elements within the pathname that can cause the pathname to resolve to a location that is outside of the restricted directory.

[Implementation] Assume all input is malicious. Use an "accept known good" input validation strategy, i.e., use a list of acceptable inputs that strictly conform to specifications. Reject any input that does not strictly conform to specifications, or transform it into something that does. When performing input validation, consider all potentially relevant properties, including length, type of input, the full range of acceptable values, missing or extra inputs, syntax, consistency across related fields, and conformance to business rules. As an example of business rule logic, "boat" may be syntactically valid because it only contains alphanumeric characters, but it is not valid if the input is only expected to contain colors such as "red" or "blue." Do not rely exclusively on looking for malicious or malformed inputs. This is likely to miss at least one undesirable input, especially if the code's environment changes. This can give attackers enough room to bypass the intended validation. However, denylis
[Architecture and Design] For any security checks that are performed on the client side, ensure that these checks are duplicated on the server side, in order to avoid CWE-602. Attackers can bypass the client-side checks by modifying values after the checks have been performed, or by changing the client to remove the client-side checks entirely. Then, these modified values would be submitted to the server.

Source: MITRE CWE corpus.