CVE-2026-9201: Langflow truncated-hash bypass enables

CISO Take

Langflow's optional hardening mode is meant to restrict custom component execution to a trusted-template allowlist, but it validates code integrity using only a truncated SHA-256 hash, and an authenticated low-privilege user can craft malicious Python component code that collides with that partial hash to slip past the check and execute arbitrary code inside the Langflow process. This isn't a cosmetic flaw — it defeats the specific control organizations enable to prevent unauthorized code execution in a framework used to build production LLM agents and pipelines, and the CVSS 8.8 score (network vector, low complexity, no user interaction, full C/I/A impact) reflects that once bypassed, compromise is complete. It's not in CISA KEV, has no public exploit or Nuclei template, and CISA's SSVC decision is TRACK — the lowest urgency tier — while EPSS sits at a low 0.23% raw probability, so there is no evidence of active or imminent exploitation. The authenticated-access prerequisite narrows the blast radius somewhat, but hardening mode exists precisely to contain lower-trust users (contractors, internal teams with limited access), so this removes the safety net for the population it was built to constrain. Track IBM's advisory for a fixed release beyond 1.10.3, and until patched, treat hardening mode as unreliable: restrict who can author or import custom components, review component code manually, and monitor for unexpected Python execution or process spawning from the Langflow service.

Sources: NVD EPSS ATLAS ibm.com

What is the risk?

High severity (CVSS 8.8) driven by full confidentiality/integrity/availability impact once exploited, network attack vector, low complexity, and no user interaction required. The mitigating factor is the PR:L requirement — exploitation needs an already-authenticated, low-privileged account, which caps this below a pre-auth RCE. Real-world exploitation likelihood currently appears low: no CISA KEV listing, no public exploit code, no scanner template, EPSS raw score of 0.23%, and CISA SSVC rates it TRACK (lowest priority tier, monitor-only). The core risk driver is that this vulnerability specifically defeats a security control (hardening mode) rather than exploiting an unprotected surface — organizations that enabled hardening mode believing it capped their exposure to trusted code are the ones most exposed by this finding.

How does the attack unfold?

Initial Access

Attacker obtains a low-privileged authenticated account on the target Langflow instance.

AML.T0012

Validation Bypass

Attacker crafts malicious custom component code whose truncated SHA-256 hash collides with a trusted template hash, defeating hardening-mode validation.

AML.T0107

Code Execution

Langflow loads and executes the malicious component, running attacker-controlled Python inside the application process.

AML.T0050

Impact

Attacker leverages process-level access to exfiltrate credentials, tamper with AI workflows, or pivot into connected systems, achieving full instance compromise.

Initial Access

Attacker obtains a low-privileged authenticated account on the target Langflow instance.

AML.T0012

Validation Bypass

Attacker crafts malicious custom component code whose truncated SHA-256 hash collides with a trusted template hash, defeating hardening-mode validation.

AML.T0107

Code Execution

Langflow loads and executes the malicious component, running attacker-controlled Python inside the application process.

AML.T0050

Impact

Attacker leverages process-level access to exfiltrate credentials, tamper with AI workflows, or pivot into connected systems, achieving full instance compromise.

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

8.8 / 10

EPSS

0.2%

chance of exploitation in 30 days

Higher than 14% 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 High

I High

A High

What should I do?

1 step

Upgrade Langflow to a fixed release above 1.10.3 as soon as IBM publishes one (track https://www.ibm.com/support/pages/node/7282646 for the patched version). Until patched, do not treat hardening mode as a sufficient control on its own: restrict which authenticated users can create or import custom components, require manual code review/approval for new components regardless of hash validation status, and apply least-privilege to all Langflow accounts. Detection: monitor the Langflow process for unexpected child processes, outbound network connections, or filesystem writes originating from component execution; audit logs for component uploads/edits by low-privileged accounts; and alert on any Python execution paths not tied to known, approved templates.

What does CISA's SSVC say?

Decision Track

Exploitation none

Automatable No

Technical Impact total

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

How is it classified?

Auth Bypass Code Execution Framework Agent Plugin AML.T0012 - Valid Accounts AML.T0050 - Command and Scripting Interpreter AML.T0107 - Exploitation for Defense Evasion

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 verification and validation

NIST AI RMF

MEASURE 2.7 - AI system security and resilience evaluation

OWASP LLM Top 10

LLM03:2025 - Supply Chain Vulnerabilities

Frequently Asked Questions

What is CVE-2026-9201?

Langflow's optional hardening mode is meant to restrict custom component execution to a trusted-template allowlist, but it validates code integrity using only a truncated SHA-256 hash, and an authenticated low-privilege user can craft malicious Python component code that collides with that partial hash to slip past the check and execute arbitrary code inside the Langflow process. This isn't a cosmetic flaw — it defeats the specific control organizations enable to prevent unauthorized code execution in a framework used to build production LLM agents and pipelines, and the CVSS 8.8 score (network vector, low complexity, no user interaction, full C/I/A impact) reflects that once bypassed, compromise is complete. It's not in CISA KEV, has no public exploit or Nuclei template, and CISA's SSVC decision is TRACK — the lowest urgency tier — while EPSS sits at a low 0.23% raw probability, so there is no evidence of active or imminent exploitation. The authenticated-access prerequisite narrows the blast radius somewhat, but hardening mode exists precisely to contain lower-trust users (contractors, internal teams with limited access), so this removes the safety net for the population it was built to constrain. Track IBM's advisory for a fixed release beyond 1.10.3, and until patched, treat hardening mode as unreliable: restrict who can author or import custom components, review component code manually, and monitor for unexpected Python execution or process spawning from the Langflow service.

Is CVE-2026-9201 actively exploited?

No confirmed active exploitation of CVE-2026-9201 has been reported, but organizations should still patch proactively.

How to fix CVE-2026-9201?

Upgrade Langflow to a fixed release above 1.10.3 as soon as IBM publishes one (track https://www.ibm.com/support/pages/node/7282646 for the patched version). Until patched, do not treat hardening mode as a sufficient control on its own: restrict which authenticated users can create or import custom components, require manual code review/approval for new components regardless of hash validation status, and apply least-privilege to all Langflow accounts. Detection: monitor the Langflow process for unexpected child processes, outbound network connections, or filesystem writes originating from component execution; audit logs for component uploads/edits by low-privileged accounts; and alert on any Python execution paths not tied to known, approved templates.

What systems are affected by CVE-2026-9201?

This vulnerability affects the following AI/ML architecture patterns: agent frameworks, LLM workflow orchestration, low-code/no-code AI builders, plugin/custom-component execution.

What is the CVSS score for CVE-2026-9201?

CVE-2026-9201 has a CVSS v3.1 base score of 8.8 (HIGH). The EPSS exploitation probability is 0.23%.

What is the AI security impact?

Affected AI Architectures

agent frameworksLLM workflow orchestrationlow-code/no-code AI buildersplugin/custom-component execution

MITRE ATLAS Techniques

AML.T0012 Valid Accounts

AML.T0050 Command and Scripting Interpreter

AML.T0107 Exploitation for Defense Evasion

Compliance Controls Affected

EU AI Act: Article 15

ISO 42001: A.6.2.6

NIST AI RMF: MEASURE 2.7

OWASP LLM Top 10: LLM03:2025

What are the technical details?

Original Advisory

IBM Langflow OSS 1.0.0 through 1.10.3 could allow an authenticated attacker to execute arbitrary code due to a cryptographic weakness in the custom component validation mechanism. When the optional hardening mode that restricts execution to trusted component templates is enabled, the application validates component code using a truncated SHA‑256 hash. Because the hash comparison relies on only a portion of the digest, an attacker can craft malicious component code that collides with a trusted template hash and bypasses validation. Successful exploitation allows the attacker to introduce and execute unauthorized Python code within the Langflow process, defeating the intended security control and potentially leading to full compromise of the affected instance.

Exploitation Scenario

An attacker obtains a low-privileged authenticated account on a Langflow instance — for example a contractor, a compromised internal user, or a limited-scope collaborator account on a shared AI-builder platform. With hardening mode enabled, the attacker crafts custom component Python code engineered to produce a truncated SHA-256 digest matching that of a trusted template hash, exploiting the fact that only a portion of the digest is compared. The manipulated component passes validation as if it were a vetted, trusted template and is loaded and executed by the Langflow engine. The attacker's code now runs with the privileges of the Langflow process itself, giving them access to any credentials, API keys, or connected services the instance manages, and a foothold for lateral movement or data exfiltration.

Weaknesses (CWE)

CWE-326 Inadequate Encryption Strength Primary CWE-326 Inadequate Encryption Strength

CWE-326 — Inadequate Encryption Strength: The product stores or transmits sensitive data using an encryption scheme that is theoretically sound, but is not strong enough for the level of protection required.