In the wild
Realized harm
CVE-2025-3248: unauthenticated RCE in Langflow's /api/v1/validate/code endpoint
AIR-2025-0044 AIR record · 2025-06-langflow-unauthenticated-rce
A real party was affected. This incident caused actual consequences, not a demonstrated capability.
From the supporting source
A remote and unauthenticated attacker can send crafted HTTP requests to execute arbitrary code.github.com
The record
Disclosed
2025-06-17
Vendor
DataStax / Langflow
Product
Langflow / langflow-base
Risk domain
Access control and permissions › Privilege escalation
System type
Generative model
Who acted
AI was the target — unauthenticated RCE on a Langflow endpoint
Attack surface
Agent framework
Autonomy level
Semi-autonomous
Initial vector
Unauthenticated http request
Impact type
Remote code execution, Botnet recruitment
Confidence
High
Submitted by
EAI Internal
Crosswalk
- OWASP Agentic Top 10 (2026)
- ASI03 – Identity and Privilege Abuse
- MITRE ATLAS
- AML.T0055 – Unsecured Credentials
- NIST AI RMF
- GOVERN 6.1
- EU AI Act
- Article 13
- ISO/IEC
- TR 24028; 42001; 23894
- AIUC-1
- B. Security
- Scope
- tool misuse
- Mitigation
- Apply least-privilege principle; Regular policy audits.
- Taxonomy
- Enkrypt AI agent risk taxonomy · Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery
Blast radius
every internet-exposed Langflow < 1.3.0; thousands of instances were reachable when mass exploitation began
Capability abused
Defense & remediation
Recorded defense
none - the endpoint required no authentication
Guardrail kinds
None
Guardrail outcome
Not applicable
Fix shipped
langflow 1.3.0 / langflow-base 0.3.0
Taxonomy & references
Curator notes
GHSA-rvqx-wpfh-mfx7, CVSS 9.3 Critical, credited to chximn-dt. Reported publicly in April 2025 and added to CISA KEV in May 2025 after mass scanning; Trend Micro later attributed a Flodrix botnet campaign to it. Those two exploitation claims could not be re-verified for this record - the CISA KEV catalog entry is not quoted here - so impact_realized true rests on widely reported mass exploitation rather than on a source cited here. Treat the KEV/Flodrix attribution as a lead to confirm. Distinct from the later CVE-2026-48519 Shareable Playground RCE and CVE-2025-57760 post-RCE superuser escalation.
Supporting sources
Cite as
Enkrypt AI. "CVE-2025-3248: unauthenticated RCE in Langflow's /api/v1/validate/code endpoint". AIR, AIR-2025-0044. https://agentincident.vercel.app/i/AIR-2025-0044/ (accessed 2026-10-19).