CVE-2026-7869: Langflow path traversal

CISO Take

IBM Langflow's Knowledge Bases API takes user-supplied knowledge base names and uses them directly to build filesystem paths, so an authenticated attacker can insert directory traversal sequences to create directories and write files anywhere on the host. Blast radius is bounded by the fact that exploitation requires a valid low-privilege account (PR:L) rather than anonymous access, and EPSS sits at 0.00216 — technically in the top 88th percentile of scored CVEs, but still a low absolute probability of near-term mass exploitation, with no public PoC, no Nuclei template, and no CISA KEV listing; CISA's own SSVC decision is TRACK, meaning routine monitoring rather than emergency action. The real risk is that a file-write primitive in a workflow/agent orchestration platform is rarely the end of the story — it's a stepping stone to overwriting application code, cron jobs, or web-servable files for persistence or RCE. Patch to the fixed version referenced in IBM's advisory as soon as it's scheduled, and in the interim restrict who can create knowledge bases, monitor `/api/v1/knowledge_bases` requests for `../` sequences in the `name` field, and run Langflow with a least-privilege, ideally read-only, filesystem.

Sources: NVD EPSS CISA KEV ATLAS ibm.com

What is the risk?

Medium severity (CVSS 5.4) reflects a real but constrained primitive: an authenticated, low-privileged user can traverse outside the intended knowledge base storage path to create arbitrary directories and files (C:L/I:L/A:N — no availability impact, limited confidentiality/integrity per the vector, though the practical ceiling is higher because arbitrary file write on a server is rarely 'limited' once chained). No active exploitation signals exist: not in CISA KEV, no public exploit code, no scanner template, and EPSS remains near-zero in absolute terms. CISA SSVC lands at TRACK, the lowest actionability tier — monitor and patch on a normal cadence rather than treat as an emergency. The requirement for authentication (PR:L) is the main mitigating factor; risk rises sharply in any multi-tenant Langflow deployment where external or lower-trust users hold accounts.

How does the attack unfold?

Authenticated Access

Attacker obtains or already holds a valid low-privileged Langflow account, satisfying the PR:L requirement.

AML.T0012

Crafted API Request

Attacker sends a POST to /api/v1/knowledge_bases with a knowledge base name containing directory traversal sequences.

AML.T0049

Arbitrary File Write

The unsanitized name is used to build the storage path, so Langflow creates directories and writes attacker-controlled files outside the intended knowledge base location.

Escalation / Persistence

Attacker leverages the file-write primitive to overwrite executable code, cron jobs, or web-accessible files, escalating toward code execution or persistent access.

Authenticated Access

Attacker obtains or already holds a valid low-privileged Langflow account, satisfying the PR:L requirement.

AML.T0012

Crafted API Request

Attacker sends a POST to /api/v1/knowledge_bases with a knowledge base name containing directory traversal sequences.

AML.T0049

Arbitrary File Write

The unsanitized name is used to build the storage path, so Langflow creates directories and writes attacker-controlled files outside the intended knowledge base location.

Escalation / Persistence

Attacker leverages the file-write primitive to overwrite executable code, cron jobs, or web-accessible files, escalating toward code execution or persistent access.

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
Langflow	pip	—	No patch
155.0K Pushed 2d ago 17% patched ~70d to patch Full package profile →

Do you use Langflow? You're affected.

How severe is it?

CVSS 3.1

5.4 / 10

EPSS

0.2%

chance of exploitation in 30 days

Higher than 12% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

No known exploitation

Sophistication

Moderate

What is the attack surface?

AV Network

AC Low

PR Low

UI None

S Unchanged

C Low

I Low

A None

What should I do?

1 step

Upgrade past Langflow 1.10.3 to the version fixed in IBM's advisory as soon as it is available and scheduled. Until patched: restrict which user roles can call the Knowledge Bases API, avoid granting broad account access in multi-tenant deployments, and add a validation layer (proxy/WAF rule or input filter) that rejects KB names containing ../, absolute paths, or other traversal patterns. Run the Langflow container/process with least-privilege filesystem permissions (ideally a read-only root filesystem with a narrowly-scoped writable data directory) so a traversal write has nowhere sensitive to land. For detection, monitor application logs and file-integrity tooling for knowledge base creation requests with path-traversal sequences in the name field, and alert on unexpected file writes outside the configured KB storage path.

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 RAG AML.T0049 - Exploit Public-Facing Application AML.T0070 - RAG Poisoning

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act

Article 15 - Accuracy, robustness and cybersecurity

OWASP LLM Top 10

LLM03 - Supply Chain Vulnerabilities

Frequently Asked Questions

What is CVE-2026-7869?

IBM Langflow's Knowledge Bases API takes user-supplied knowledge base names and uses them directly to build filesystem paths, so an authenticated attacker can insert directory traversal sequences to create directories and write files anywhere on the host. Blast radius is bounded by the fact that exploitation requires a valid low-privilege account (PR:L) rather than anonymous access, and EPSS sits at 0.00216 — technically in the top 88th percentile of scored CVEs, but still a low absolute probability of near-term mass exploitation, with no public PoC, no Nuclei template, and no CISA KEV listing; CISA's own SSVC decision is TRACK, meaning routine monitoring rather than emergency action. The real risk is that a file-write primitive in a workflow/agent orchestration platform is rarely the end of the story — it's a stepping stone to overwriting application code, cron jobs, or web-servable files for persistence or RCE. Patch to the fixed version referenced in IBM's advisory as soon as it's scheduled, and in the interim restrict who can create knowledge bases, monitor `/api/v1/knowledge_bases` requests for `../` sequences in the `name` field, and run Langflow with a least-privilege, ideally read-only, filesystem.

Is CVE-2026-7869 actively exploited?

No confirmed active exploitation of CVE-2026-7869 has been reported, but organizations should still patch proactively.

How to fix CVE-2026-7869?

Upgrade past Langflow 1.10.3 to the version fixed in IBM's advisory as soon as it is available and scheduled. Until patched: restrict which user roles can call the Knowledge Bases API, avoid granting broad account access in multi-tenant deployments, and add a validation layer (proxy/WAF rule or input filter) that rejects KB names containing `../`, absolute paths, or other traversal patterns. Run the Langflow container/process with least-privilege filesystem permissions (ideally a read-only root filesystem with a narrowly-scoped writable data directory) so a traversal write has nowhere sensitive to land. For detection, monitor application logs and file-integrity tooling for knowledge base creation requests with path-traversal sequences in the name field, and alert on unexpected file writes outside the configured KB storage path.

What systems are affected by CVE-2026-7869?

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

What is the CVSS score for CVE-2026-7869?

CVE-2026-7869 has a CVSS v3.1 base score of 5.4 (MEDIUM). The EPSS exploitation probability is 0.22%.

What is the AI security impact?

Affected AI Architectures

agent frameworksRAG pipelinesLLM orchestration platforms

MITRE ATLAS Techniques

AML.T0049 Exploit Public-Facing Application

AML.T0070 RAG Poisoning

Compliance Controls Affected

EU AI Act: Article 15

OWASP LLM Top 10: LLM03

What are the technical details?

Original Advisory

IBM Langflow OSS 1.0.0 through 1.10.3 is vulnerable to Path Traversal in the Knowledge Bases API (`POST /api/v1/knowledge_bases`). This occurs because user-supplied knowledge base names are used directly to create file paths without proper sanitization or containment checks. An authenticated attacker can exploit this flaw to create directories and write files anywhere on the server's filesystem.

Exploitation Scenario

An attacker who holds a valid but low-privileged Langflow account (e.g., a standard contributor in a shared or multi-tenant instance) sends a `POST /api/v1/knowledge_bases` request with a knowledge base name like `../../../../opt/app/scripts/startup.py`. Because the name is used directly to build the storage path with no containment check, Langflow creates the directory structure and writes the attacker-supplied content outside the intended knowledge base folder — for example overwriting an application script, planting a file in a location loaded on the next process restart, or dropping a web-servable file into a reachable web root. What begins as a medium-severity authenticated file-write bug becomes a path to persistence or remote code execution once chained with how the written file is subsequently loaded or executed.

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.