Is my AI tool safe?Agent frameworks and buildersLangflow (DataStax, IBM)
Is Langflow safe?
No tool is safe or unsafe on its own: it depends on what it can reach and how it is set up. The record for Langflow: 5 incidents in this center's library and 12 CVEs in its advisory feed (7 critical, 5 high), the newest published Oct 6, 2026. Below: what went wrong, what was fixed, and how to use it safely.
What to watch for
The clearest case of an AI tool that must never face the internet: an unauthenticated code-execution endpoint, exploited within a day, listed by CISA as known-exploited, and the entry point for an agent-run ransomware attack a year later.
How to use it safely
Keep it behind authentication and off the public internet, update past 1.3.0 and stay current, and remember it holds the API keys of every flow built in it.
What it can reach
- Often exposed to the internet
- Runs commands with your permissions
- Holds API keys and tokens
What went wrong: incidents in the library
- Langflow: a logged-in user's MCP server command runs as a shell, and honeypots have counted 2,300 exploit attempts since September Critical
CVE-2026-105740 is an authenticated remote code execution flaw in Langflow, rated CVSS 9.9 in the GitHub advisory GHSA-7w94-79vh-5mr2 that Langflow published on September 10, 2026: any logged-in user can run operating system… - Langflow: IBM lists more than two dozen flaws, two of them code execution with no login, and an LLM-written lambda that runs on the server Critical
IBM's security bulletin of October 2, "Langflow OSS is affected by multiple vulnerabilities", covers Langflow OSS 1.0.0 through 1.12.2 and lists more than two dozen CVEs, fixed in 1.12.3. Two are rated 9.8 and need no login:… - An open-source agent and DeepSeek run hundreds of attacks for one operator High
Unit 42 documented a single operator who used the open-source Hermes Agent framework with DeepSeek as its reasoning model to find internet-facing systems, pick public exploits and attack them with little further input. It went… - JADEPUFFER: an AI agent runs a ransomware attack from Langflow to database Critical
Sysdig's threat research team documented an attack in which an AI agent carried out the whole chain: it broke into an exposed Langflow server, harvested credentials, moved to a production database server, encrypted 1,342… - Langflow: unauthenticated code execution exploited in the wild Critical
Langflow, a popular visual builder for AI apps, had an API endpoint that ran Python code from anyone on the internet without a login. Attackers exploited it, CISA listed it as known exploited, and one campaign used it to build a…
Known CVEs
- CVE-2026-10561 Critical Langflow Python interpreter components run unsandboxed code, letting a logged-in user take over
- CVE-2026-9205 Critical Langflow derives the key that encrypts stored credentials from a predictable random seed
- CVE-2026-7700 High Langflow Smart Transform runs the lambda an LLM writes, so prompt injection becomes code execution
- CVE-2026-105697 Critical Langflow runs any command placed in an MCP stdio server configuration
- CVE-2026-105699 High Langflow MCP resource handler reads files from other users' flows
- CVE-2026-105741 High Langflow trusts a spoofed X-Forwarded-For header and lets a remote user write IDE configuration
- CVE-2026-8505 Critical Langflow unauthenticated flow execution via webhook authentication bypass
- CVE-2026-51886 High Langflow authenticated code execution via decorators in validate_code
- CVE-2026-33017 Critical Known exploited Langflow unauthenticated RCE via the public flow build endpoint
- CVE-2025-34291 Critical Langflow CORS misconfiguration enables account takeover and RCE
- CVE-2025-57760 High Langflow privilege escalation via CLI superuser creation
- CVE-2025-3248 Critical Known exploited Langflow unauthenticated remote code execution
How many people use it
- 154k GitHub stars · GitHub
- 40k PyPI downloads in a month (to Oct 1, 2026) · PyPI Stats
A high CVE count can mean heavy scrutiny as much as weak engineering, so read the incidents and advisories, not only the number. The counts are what this center has recorded, not every flaw ever found; usage figures are the vendors' and registries' own published numbers.
Sources
- GitHub: GitHub stars
- PyPI Stats: PyPI downloads in a month (to Oct 1, 2026)
- GitHub Advisory Database: langflow
- Langflow: GitHub advisory GHSA-7w94-79vh-5mr2, authenticated RCE via MCP stdio transport
- Previdian: CVE-2026-105740 alert and honeypot telemetry
- The Hacker Wire: CVE-2026-105697
- IBM: Security Bulletin: Langflow OSS is affected by multiple vulnerabilities
- GitHub advisory GHSA-9fpm-3445-2vx4: prompt injection in Langflow Smart Transform can lead to code execution
- GitHub advisory GHSA-8qpj-27x8-pwpq: Langflow Python interpreter runs unsandboxed code
- GitHub advisory GHSA-jxw3-mjmx-3pqm: Langflow weak Fernet key via random.seed()
- Unit 42: autonomous AI cyberattack campaign
- The Hacker News: hacker commands DeepSeek through an agent
- Sysdig: JADEPUFFER, agentic ransomware for automated database extortion
- Horizon3.ai: unauthenticated RCE in Langflow
- Trend Micro: Langflow exploited to deliver the Flodrix botnet
- The Hacker News: Langflow flaw added to CISA KEV