Loading vulnerability…
CVE-2026-17632 is an authenticated remote code execution vulnerability in IBM Langflow OSS affecting versions 1.0.0 through 1.10.3. The flaw is caused by improper validation of Python code during AST-based security scanning. An authenticated attacker can supply crafted Python code that bypasses or defeats the intended validation logic in the AST-based scanning process, resulting in execution of arbitrary code on the target system. The issue is classified as CWE-94.
Mallory correlates every CVE against your assets, your vendors, and active adversary campaigns. Know which vulnerabilities matter for you, not just which ones are loud.
What it means. What to do now. Patch path, mitigations, and the assume-compromise checklist.
What an attacker gets, and what they’ve been doing with it.
If you can’t patch tonight, do this now.
Patch, then assume compromise.
1 valid exploit after Mallory filtered fakes, detection scripts, and README-only repos.
This three-file repository is an operational Python proof of concept for the claimed CVE-2026-17633 authenticated RCE in Langflow OSS. Its primary entry point, exploit_CVE-2026-17633.py, performs a GET health check and then sends an authenticated JSON POST containing generated Python source to /api/v1/custom_component. The generated component imports os and executes an operator-selected shell command in the component class body. The stated rationale is that class-body code is retained and executed during Langflow custom-component compilation, whereas a module-level expression may be discarded by the described preprocessing path. HTTP 200 is treated as confirmation that the component was processed and the command triggered, although the script does not independently retrieve or verify command output. The exploit requires a bearer token and the target setting LANGFLOW_ALLOW_CUSTOM_COMPONENTS=true; it reports common HTTP failure conditions for invalid authentication, disabled custom components, and missing endpoints. Command execution is configurable with -c/--command, with id > /tmp/pwned.txt as the default payload. docker-compose.yml supplies a reproducible local Langflow 1.10.3 lab configuration, including a loopback-only 127.0.0.1:7860 port binding, explicit test credentials, disabled SSRF protection, agentic experience enabled, and custom components enabled. README.md documents both this primary exploit and a separate claimed CVE-2026-17632 AST security-scanner bypass involving the Agentic Assistant flow; however, no code in the repository automates or delivers that second vector. No recognized exploit framework is used.
Products and vendors Mallory has correlated with this vulnerability. Open in Mallory to drill down to specific CPE configurations and version ranges.
Vendor-confirmed product mapping. Mallory continuously reconciles this list against your asset inventory.
4 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.
Query your assets running an affected version, and investigate the blast radius.
Every observed campaign linking this CVE to a named adversary.
Malware families riding this exploit, with evidence and IOCs.
YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Cross-references every affected SKU, including bundled OEM variants.
Community discussion across Reddit, Mastodon, and other social sources.