CVE-2026-7700: Langflow Smart Transform evals the lambda an LLM writes, so prompt injection can run code (fixed 1.10.3)
Langflow's Smart Transform component (LambdaFilterComponent) asks an LLM to write a Python lambda from the flow author's instructions and a preview of the input, checks only that the reply starts with 'lambda' and contains ':', and then eval()s it with full builtins inside the Langflow process. A flow author can run code directly, and attacker-controlled data passed into an exposed flow can do it through prompt injection. Langflow 1.3.0 to 1.10.2 are affected; fixed in 1.10.3 and 1.11.0.
What it is
Vulnerable code: src/lfx/src/lfx/components/llm_operations/lambda_filter.py, where _validate_lambda() only checks the prefix and a colon before eval(lambda_text) (CWE-94).
A reply like lambda x: __import__("os").system("id") passes the check and runs when the component applies the function to the data, with the privileges of the Langflow service.
Direct route: anyone who can edit a flow's Instructions field. Indirect route: content an outside user controls flowing into Smart Transform in a public or API-exposed flow, if the model follows the injected instruction.
The fix validates the generated code's AST and evaluates it with restricted builtins (PR #13530 in 1.11.0, backported to 1.10.3 in PR #14071). Reported by Peyton Kennedy of Endor Labs; GitHub published the advisory to its database on 5 October 2026.
How to check
- Check your Langflow version (`pip show langflow` or the version in the UI); 1.3.0 through 1.10.2 are affected.
- Search your flows for the Smart Transform component and note which ones take input from users, webhooks or fetched documents.
How to fix
- Upgrade to Langflow 1.10.3 or later (1.11.0+ on the main line).
- Until then, remove Smart Transform from runnable flows, keep untrusted data out of it, and restrict who can create or edit flows.
- Run Langflow with minimal filesystem, network and credential access.
pip show langflow | grep -i ^version
# find flows that use Smart Transform in an exported flows folder
grep -rl 'LambdaFilterComponent' ./flows 2>/dev/null Sources
- criticalCVE-2026-93680 and CVE-2026-110186: IBM fixes a 9.8 authentication bypass and unauthenticated code execution in Langflow up to 1.12.8, upgrade to 1.13.0
- criticalCVE-2026-104339 and 30 more: IBM's Langflow 1.12.4 and 1.12.5 bulletins fix unauthenticated RCE, three 9.9 code execution bugs, SSRF and path traversal
- criticalCVE-2026-104334 and CVE-2026-93674: IBM's Langflow OSS bulletin lists two unauthenticated code execution flaws plus a batch of sandbox, file access and credential issues, fixed in 1.12.3
- criticalCVE-2026-10561: Langflow's Python Interpreter component ran code with no real sandbox