Loading story…
A critical vulnerability tracked as CVE-2026-9198 allows unauthenticated remote code execution against default-configured Langflow OSS deployments by chaining two exposed API behaviors. According to IBM's security bulletin, the /api/v1/auto_login endpoint can issue SUPERUSER bearer tokens to any network caller, and the /api/v1/validate/code endpoint then executes attacker-controlled Python through exec(). IBM said the validator can trigger execution through Python decorators, default arguments, and annotations at function definition time, giving attackers arbitrary command execution on the host. The issue affects Langflow OSS versions 1.0.0 through 1.10.0, carries a CVSS 3.1 score of 9.8, and has no listed workaround.
A draft Metasploit Framework pull request adds an exploit module for the flaw and says successful exploitation can yield a Meterpreter session on vulnerable systems. The module author reported testing against Langflow 1.8.4 on Ubuntu 22.04, describing the attack as unauthenticated and effective on default deployments. The pull request also links the bug to CISA KEV, underscoring active operational relevance, while IBM advises organizations to upgrade to Langflow OSS 1.10.1 to remediate the exposure.
Mallory correlates global threat intelligence with your attack surface — know if you’re exposed before adversaries strike.
4 events from the most recent confirmed update back to the earliest known activity.
On 2026-08-04, a GitHub automation bot added Rapid7 Metasploit pull request #21753 to the Metasploit Kanban project and moved it to Todo. The pull request was the draft exploit module for CVE-2026-9198.
IBM, acting as the CNA, published the CVE record for CVE-2026-9198, formally documenting the critical Langflow OSS remote code execution flaw affecting versions 1.0.0 through 1.10.0. The record described the exploit chain involving /api/v1/auto_login and /api/v1/validate/code and assigned the issue a CVSS 9.8 severity rating.
On 2026-07-02, IBM published a security bulletin for CVE-2026-9198, describing an unauthenticated remote code execution flaw in default-configured Langflow OSS deployments. The bulletin said the issue chains the /api/v1/auto_login endpoint with /api/v1/validate/code and recommended upgrading to Langflow OSS 1.10.1.
A draft pull request in Rapid7's Metasploit Framework proposed a new exploit module for CVE-2026-9198, stating the chain can yield full unauthenticated RCE against default Langflow deployments and was tested against Langflow 1.8.4 on Ubuntu 22.04. The PR also noted that successful exploitation results in a Meterpreter session.
Vulnerabilities, threat actors, malware, products, organizations, and breaches Mallory has linked to this story.
3 references tracked. Mallory keeps watching after this page renders.
github.com
Open sourcecve.org
Open sourceibm.com
Open sourceMap indicators from this story to your assets and identify affected systems in minutes.
Every observed campaign, victim, and pivot linked to actors named in this story.
Malware, exploits, and IOCs connected to the activity described here.
YARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Get matching new stories delivered to your team as they break — not the next morning.
Ask questions about this story and take action on the answers.
Other stories Mallory is tracking that overlap on entities, sources, or topic labels.
IBM disclosed **CVE-2026-10561**, a critical unauthenticated remote code execution flaw in Langflow OSS affecting versions `1.0.0` through `1.9.3`. The vulnerability is rooted in improper isolation in the `PythonREPLComponent`: CPython's `exec()` restores full builtins access when the globals dictionary does not explicitly clear `__builtins__`, allowing attackers to bypass the intended import restrictions. IBM and downstream advisories rate the issue **CVSS 3.1 10.0**, noting it is remotely exploitable with low attack complexity, requires no privileges, and needs no user interaction. The flaw can be chained with Langflow's default `LANGFLOW_AUTO_LOGIN=true` behavior, which exposes the `/api/v1/auto_login` endpoint and can issue a superuser JWT without credentials, enabling unauthenticated code execution on the host. Successful exploitation could lead to arbitrary OS command execution, theft of LLM provider keys, flow definitions and vector store credentials, and persistent compromise of affected systems. IBM recommends upgrading to **Langflow OSS 1.9.4** immediately; no workaround was listed, and defenders are advised to apply vendor updates, harden access controls, and monitor for unauthorized activity.
Jul 12, 2026
IBM disclosed multiple severe vulnerabilities in **Langflow OSS** affecting versions `1.0.0` through `1.10.1`, including remote code execution, denial of service, path traversal, unauthorized file and vector store access, build job information disclosure, and exposed credentials through unauthenticated or insufficiently authorized API endpoints. The most prominent issue, **`CVE-2026-13446`**, is a hard-coded credentials flaw classified as **CWE-798** and scored **CVSS 9.8**, creating a network-exploitable path to compromise confidentiality, integrity, and availability. According to IBM's bulletin, the vulnerable endpoints could let attackers execute code, read arbitrary files, access secrets, modify other users' data, and cancel active builds without proper authorization. IBM advised users to upgrade Langflow OSS to **`1.10.2`** to remediate the issues, as the affected releases expose both application internals and connected external components to critical compromise.
Jul 18, 2026
IBM disclosed five high-severity vulnerabilities in **Langflow OSS** affecting versions `1.0.0` through `1.10.0`, including multiple paths to **remote code execution**, privilege escalation, and arbitrary file write. The issues are tracked as `CVE-2026-9135`, `CVE-2026-8476`, `CVE-2026-8481`, `CVE-2026-8635`, and `CVE-2026-8859`. Reported weaknesses include code injection through unvalidated dynamic `CodeInput` fields in the Policies component, unsafe `pickle.loads()` deserialization in the disk cache, direct execution of user-supplied Python in the `/api/v1/validate/code` endpoint, database manipulation that can elevate an authenticated user to superuser, and path traversal in the APIRequest component's **Save to File** feature via crafted `Content-Disposition` filenames. Several of the flaws require only authenticated access and can lead to arbitrary Python or system command execution with the privileges of the Langflow server process, creating a path to full host compromise in some deployments. IBM released **Langflow OSS `1.10.1`** to address all five vulnerabilities and advised organizations to upgrade immediately; additional mitigation guidance includes reviewing file-system permissions and sanitizing or removing cached data where possible. Public reporting said no active exploitation had been confirmed at publication time.
Jul 19, 2026