AI Threat AlertCVE-2026-18899: Langflow: path traversal enables arbitrary file read
HIGHIBM Langflow OSS versions 1.0.0 through 1.11.1 contain a path traversal flaw (CWE-22) that lets a remote, unauthenticated attacker read arbitrary files from the host filesystem over the network. This matters because Langflow instances typically hold .env files, LLM provider API keys, database credentials, and flow definitions on disk — exactly the kind of secrets a low-complexity, no-auth, no-interaction bug (CVSS 7.5, AV:N/AC:L/PR:N/UI:N) is well-suited to exfiltrate. There is no EPSS score, no CISA KEV listing, and no public exploit or Nuclei template as of this writing, so there is no evidence yet of active exploitation, but the attack pattern (an unauthenticated network-reachable file-read primitive) is exactly what opportunistic scanners weaponize quickly once a PoC surfaces. Patch any Langflow deployment in the affected range to the fixed release immediately, and until then ensure the instance is not reachable from untrusted networks. Rotate any credentials or API keys that were stored on a Langflow host prior to patching, and add detection for path traversal sequences (../, URL-encoded variants) in Langflow access logs.
What is the risk?
High severity (CVSS 7.5) driven by a confidentiality-only impact, but the combination of network vector, low attack complexity, and zero privileges/interaction required makes this trivially exploitable by any actor who can reach the service. Absence of EPSS data, KEV listing, public exploit code, or a Nuclei template means real-world exploitation is currently unconfirmed, but path traversal bugs in self-hosted AI tooling are commonly and quickly automated once disclosed, especially given Langflow's history of internet-exposed, low-code deployments. Organizations running affected versions reachable from the internet or from untrusted internal segments should treat this as urgent-to-patch even without confirmed in-the-wild activity.
How does the attack unfold?
What systems are affected?
| Package | Ecosystem | Vulnerable Range | Patched |
|---|---|---|---|
| Langflow | pip | — | No patch |
Do you use Langflow? You're affected.
How severe is it?
What is the attack surface?
What should I do?
1 step1) Upgrade Langflow to the version that remediates this issue (beyond 1.11.1) as soon as it is available and verified — check IBM's advisory (https://www.ibm.com/support/pages/node/7284579) for the exact fixed release. 2) Until patched, do not expose Langflow instances directly to the internet; place them behind a VPN, reverse proxy with authentication, or restrict access via firewall/security group to known IPs. 3) Add WAF/reverse-proxy rules to block path traversal patterns (../, ..%2f, encoded variants) targeting Langflow endpoints. 4) Audit and rotate any secrets (API keys, DB credentials, .env contents) that resided on a Langflow host, since you cannot confirm retroactively whether they were read. 5) Review Langflow access logs for anomalous file-path parameters or repeated 200/404 responses on file-serving routes as an indicator of prior exploitation attempts. 6) Ensure Langflow runs with least-privilege filesystem permissions so even a successful traversal has a narrower blast radius.
What does CISA's SSVC say?
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:
Frequently Asked Questions
What is CVE-2026-18899?
IBM Langflow OSS versions 1.0.0 through 1.11.1 contain a path traversal flaw (CWE-22) that lets a remote, unauthenticated attacker read arbitrary files from the host filesystem over the network. This matters because Langflow instances typically hold .env files, LLM provider API keys, database credentials, and flow definitions on disk — exactly the kind of secrets a low-complexity, no-auth, no-interaction bug (CVSS 7.5, AV:N/AC:L/PR:N/UI:N) is well-suited to exfiltrate. There is no EPSS score, no CISA KEV listing, and no public exploit or Nuclei template as of this writing, so there is no evidence yet of active exploitation, but the attack pattern (an unauthenticated network-reachable file-read primitive) is exactly what opportunistic scanners weaponize quickly once a PoC surfaces. Patch any Langflow deployment in the affected range to the fixed release immediately, and until then ensure the instance is not reachable from untrusted networks. Rotate any credentials or API keys that were stored on a Langflow host prior to patching, and add detection for path traversal sequences (../, URL-encoded variants) in Langflow access logs.
Is CVE-2026-18899 actively exploited?
No confirmed active exploitation of CVE-2026-18899 has been reported, but organizations should still patch proactively.
How to fix CVE-2026-18899?
1) Upgrade Langflow to the version that remediates this issue (beyond 1.11.1) as soon as it is available and verified — check IBM's advisory (https://www.ibm.com/support/pages/node/7284579) for the exact fixed release. 2) Until patched, do not expose Langflow instances directly to the internet; place them behind a VPN, reverse proxy with authentication, or restrict access via firewall/security group to known IPs. 3) Add WAF/reverse-proxy rules to block path traversal patterns (../, ..%2f, encoded variants) targeting Langflow endpoints. 4) Audit and rotate any secrets (API keys, DB credentials, .env contents) that resided on a Langflow host, since you cannot confirm retroactively whether they were read. 5) Review Langflow access logs for anomalous file-path parameters or repeated 200/404 responses on file-serving routes as an indicator of prior exploitation attempts. 6) Ensure Langflow runs with least-privilege filesystem permissions so even a successful traversal has a narrower blast radius.
What systems are affected by CVE-2026-18899?
This vulnerability affects the following AI/ML architecture patterns: agent frameworks, RAG pipelines.
What is the CVSS score for CVE-2026-18899?
CVE-2026-18899 has a CVSS v3.1 base score of 7.5 (HIGH). The EPSS exploitation probability is 0.69%.
What is the AI security impact?
Affected AI Architectures
MITRE ATLAS Techniques
AML.T0037Data from Local SystemAML.T0049Exploit Public-Facing ApplicationAML.T0055Unsecured CredentialsCompliance Controls Affected
What are the technical details?
Original Advisory
IBM Langflow OSS 1.0.0 through 1.11.1 could allow a remote attacker to read arbitrary files due to path traversal.
Exploitation Scenario
An attacker scans the internet or an internal network for exposed Langflow instances (a distinctive, recognizable low-code AI UI and API surface). Finding one on the affected version, they craft an HTTP request against a file-related endpoint using traversal sequences (e.g., ../../../.env or encoded equivalents) with no authentication required. The server returns the contents of the targeted file, which may include the Langflow .env with LLM provider API keys, database credentials, or other configuration secrets. The attacker then uses the harvested API keys to run up usage costs on the victim's LLM provider account, access connected vector databases or RAG data sources, or pivot into other systems that share those credentials.
Weaknesses (CWE)
CWE-22Improper Limitation of a Pathname to a Restricted DirectoryPrimaryCWE-22Improper Limitation of a Pathname to a Restricted DirectoryCWE-22 — Improper Limitation of a Pathname to a Restricted Directory: 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.
CVSS Vector
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:NReferences
Timeline
Related Vulnerabilities
CVE-2026-1056110.0Langflow: auth bypass + unauthenticated RCE (CVSS 10)
Same package: langflowCVE-2026-1013410.0Langflow: unauthenticated RCE via tool_code injection
Same package: langflowCVE-2026-333099.9langflow: Path Traversal enables file access
Same package: langflowCVE-2026-552559.9Langflow: IDOR allows cross-user flow execution
Same package: langflowCVE-2026-78739.9Langflow: authenticated RCE enables credential theft
Same package: langflow