CVE-2026-3340: IBM Langflow: SSRF enables internal network enumeration

MEDIUM
Published April 30, 2026
CISO Take

IBM Langflow Desktop versions 1.0.0 through 1.8.4 contain an SSRF flaw that coerces the application server into issuing unauthorized HTTP requests to internal network resources on behalf of an attacker. Despite a medium CVSS score of 6.5, EPSS data places this in the top 91st percentile for exploitation likelihood — a signal that real-world conditions favor exploitation faster than severity alone suggests, and the CVSS vector (PR:N, AC:L) contradicts the advisory's 'authenticated attacker' qualifier, warranting treatment as potentially pre-auth. Langflow's architectural position as an LLM workflow orchestrator — typically wired to internal model endpoints, vector databases, and cloud APIs — amplifies the blast radius well beyond a standard SSRF: cloud metadata endpoints (AWS/GCP/Azure IMDS), internal credential stores, and connected AI services are all reachable. Update to the remediated release per the IBM advisory and apply strict egress filtering on the Langflow host to block RFC1918, link-local, and loopback targets pending patching.

Sources:NVDEPSSATLASibm.com

What is the risk?

Moderate-to-elevated risk for any organization running Langflow adjacent to cloud infrastructure or internal AI services. The CVSS attack complexity is low and the attack vector is network, meaning no special positioning is required. The discrepancy between the CVSS vector (PR:N — no privileges) and the advisory description ('authenticated attacker') is unresolved; organizations should conservatively assume the flaw may be reachable without authentication. EPSS top-91st percentile indicates elevated exploitation probability relative to the broader CVE corpus. Absence from CISA KEV and lack of a public exploit or Nuclei template reduce immediate urgency, with SSVC decision TRACK appropriate.

How does the attack unfold?

Initial Access
Attacker accesses IBM Langflow Desktop (v1.0.0–1.8.4) via an authenticated session or potentially without credentials given the CVSS PR:N vector discrepancy with the advisory description.
AML.T0012
Exploitation
Attacker crafts a malicious Langflow workflow containing an HTTP Request node targeting internal IPs, cloud metadata endpoints (169.254.169.254), or internal AI service URLs.
AML.T0049
Discovery
The Langflow server issues unauthorized outbound HTTP requests, returning responses that reveal internal host topology, cloud IAM credentials from IMDS, or connected AI service data.
AML.T0006
Impact
Attacker leverages harvested cloud credentials or enumerated internal endpoints to pivot into AI infrastructure — model APIs, vector databases, training data stores — escalating well beyond the Langflow instance itself.
AML.T0075

What systems are affected?

PackageEcosystemVulnerable RangePatched
Langflowpip—No patch
156.7KPushed 3d ago18% patched~76d to patchFull package profile →

Do you use Langflow? You're affected.

How severe is it?

CVSS 3.1
6.5 / 10
EPSS
0.2%
chance of exploitation in 30 days
Higher than 6% of all CVEs
Exploitation Status
No known exploitation
Sophistication
Trivial

What is the attack surface?

AVACPRUISCIA
AVNetwork
ACLow
PRNone
UINone
SUnchanged
CLow
ILow
ANone

What should I do?

5 steps
  1. Patch: Apply the remediated IBM Langflow Desktop version per the advisory at https://www.ibm.com/support/pages/node/7271096 immediately.

  2. Egress filtering: Enforce strict outbound HTTP allowlisting on the Langflow host — deny requests to RFC1918 ranges (10.x, 172.16-31.x, 192.168.x), link-local (169.254.x.x), and loopback (127.x).

  3. Network segmentation: Isolate Langflow instances from sensitive internal AI services, model APIs, and cloud control planes using firewall rules or service mesh policy.

  4. Detection: Alert on outbound HTTP requests from the Langflow process to internal IP ranges, cloud metadata addresses, or unexpected external hosts.

  5. Authentication enforcement: Given the CVSS PR:N discrepancy, verify authentication is enforced on all exposed Langflow instances and disable anonymous access if present.

What does CISA's SSVC say?

DecisionTrack
Exploitationnone
AutomatableYes
Technical Impactpartial

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

How is it classified?

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act
Article 9-Risk management system
ISO 42001
A.6.2-AI system security — network access controls
NIST AI RMF
MANAGE 2.2-Mechanisms to sustain the trustworthiness of AI systems over the operational lifecycle
OWASP LLM Top 10
LLM06-Excessive Agency

Frequently Asked Questions

What is CVE-2026-3340?

IBM Langflow Desktop versions 1.0.0 through 1.8.4 contain an SSRF flaw that coerces the application server into issuing unauthorized HTTP requests to internal network resources on behalf of an attacker. Despite a medium CVSS score of 6.5, EPSS data places this in the top 91st percentile for exploitation likelihood — a signal that real-world conditions favor exploitation faster than severity alone suggests, and the CVSS vector (PR:N, AC:L) contradicts the advisory's 'authenticated attacker' qualifier, warranting treatment as potentially pre-auth. Langflow's architectural position as an LLM workflow orchestrator — typically wired to internal model endpoints, vector databases, and cloud APIs — amplifies the blast radius well beyond a standard SSRF: cloud metadata endpoints (AWS/GCP/Azure IMDS), internal credential stores, and connected AI services are all reachable. Update to the remediated release per the IBM advisory and apply strict egress filtering on the Langflow host to block RFC1918, link-local, and loopback targets pending patching.

Is CVE-2026-3340 actively exploited?

No confirmed active exploitation of CVE-2026-3340 has been reported, but organizations should still patch proactively.

How to fix CVE-2026-3340?

1. Patch: Apply the remediated IBM Langflow Desktop version per the advisory at https://www.ibm.com/support/pages/node/7271096 immediately. 2. Egress filtering: Enforce strict outbound HTTP allowlisting on the Langflow host — deny requests to RFC1918 ranges (10.x, 172.16-31.x, 192.168.x), link-local (169.254.x.x), and loopback (127.x). 3. Network segmentation: Isolate Langflow instances from sensitive internal AI services, model APIs, and cloud control planes using firewall rules or service mesh policy. 4. Detection: Alert on outbound HTTP requests from the Langflow process to internal IP ranges, cloud metadata addresses, or unexpected external hosts. 5. Authentication enforcement: Given the CVSS PR:N discrepancy, verify authentication is enforced on all exposed Langflow instances and disable anonymous access if present.

What systems are affected by CVE-2026-3340?

This vulnerability affects the following AI/ML architecture patterns: LLM workflow orchestration platforms, agent frameworks, AI development environments, cloud-hosted AI pipelines.

What is the CVSS score for CVE-2026-3340?

CVE-2026-3340 has a CVSS v3.1 base score of 6.5 (MEDIUM). The EPSS exploitation probability is 0.17%.

What is the AI security impact?

Affected AI Architectures

LLM workflow orchestration platformsagent frameworksAI development environmentscloud-hosted AI pipelines

MITRE ATLAS Techniques

AML.T0006Active Scanning
AML.T0037Data from Local System
AML.T0049Exploit Public-Facing Application
AML.T0075Cloud Service Discovery

Compliance Controls Affected

EU AI Act: Article 9
ISO 42001: A.6.2
NIST AI RMF: MANAGE 2.2
OWASP LLM Top 10: LLM06

What are the technical details?

Original Advisory

IBM Langflow Desktop 1.0.0 through 1.8.4 IBM Langflow is vulnerable to server-side request forgery (SSRF). This may allow an authenticated attacker to send unauthorized requests from the system, potentially leading to network enumeration or facilitating other attacks.

Exploitation Scenario

An attacker — authenticated or potentially unauthenticated given the PR:N CVSS vector — accesses an internet-facing IBM Langflow Desktop instance and creates a flow containing an HTTP Request node targeting http://169.254.169.254/latest/meta-data/iam/security-credentials/ (AWS IMDS). Langflow's backend processes the flow and returns the cloud IAM temporary credentials to the attacker. With these credentials the attacker pivots directly into the cloud account hosting the AI stack: accessing S3 buckets containing training data, calling model inference APIs under the victim's identity, or enumerating the internal VPC to find vector databases and other AI services that lack public authentication.

Weaknesses (CWE)

CWE-918 — Server-Side Request Forgery: The web server receives a URL or similar request from an upstream component and retrieves the contents of this URL, but it does not sufficiently ensure that the request is being sent to the expected destination.

  • [Implementation, Operation] If there is a known list of applications, IP, hostname, and/or domains to which the product is allowed to communicate, then use allowlists to ensure that outgoing requests are restricted to these known entities. The allowlists could be in the application code or possibly enforced by firewalls or network segmentation [REF-1511].
  • [Implementation, Operation] If the product is expected to be able to connect to arbitrary IP addresses or domains, consider using a denylist that includes private network addresses, internal domain names, etc. The denylist could be in the application code or possibly enforced by firewalls or network segmentation [REF-1511].

Source: MITRE CWE corpus.

CVSS Vector

CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:L/A:N

Timeline

Published
April 30, 2026
Last Modified
May 11, 2026
First Seen
April 30, 2026

Related Vulnerabilities