CVE-2025-3248

Published Apr 7, 2025

Last updated a month ago

Overview

AI description

Automated description summarized from trusted sources.

CVE-2025-3248 is a code injection vulnerability that affects Langflow versions prior to 1.3.0. It exists in the `/api/v1/validate/code` endpoint, where a remote, unauthenticated attacker can send crafted HTTP requests to execute arbitrary code on the server. This vulnerability allows attackers to gain control of vulnerable Langflow servers without needing authentication. To remediate this vulnerability, users are advised to upgrade to Langflow version 1.3.0 or restrict network access to the application.

Description
Langflow versions prior to 1.3.0 are susceptible to code injection in the /api/v1/validate/code endpoint. A remote and unauthenticated attacker can send crafted HTTP requests to execute arbitrary code.
Source
disclosure@vulncheck.com
NVD status
Analyzed
Products
langflow

Risk scores

CVSS 3.1

Type
Primary
Base score
9.8
Impact score
5.9
Exploitability score
3.9
Vector string
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
Severity
CRITICAL

Known exploits

Data from CISA

Vulnerability name
Langflow Missing Authentication Vulnerability
Exploit added on
May 5, 2025
Exploit action due
May 26, 2025
Required action
Apply mitigations per vendor instructions, follow applicable BOD 22-01 guidance for cloud services, or discontinue use of the product if mitigations are unavailable.

Weaknesses

disclosure@vulncheck.com
CWE-306
nvd@nist.gov
CWE-94

Social media

Hype score
Not currently trending
  1. Researchers observed autonomous AI agents exploiting default credentials and CVE-2025-3248 across internet-exposed LiteLLM, Dify, and Langflow deployments — harvesting API keys at over 69 requests per minute and immediately reusing stolen credentials to fund further

    @aviatrixtrc

    16 Nov 2026

    190 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  2. Amazon: China-linked operators' AI agents hunted exposed AI apps (LiteLLM, Dify, Langflow, RAGFlow) to steal stored AI provider keys. Playbook: default creds, leaked secrets, Langflow CVE-2025-3248. Change defaults, patch, rotate keys. #AISecurity https://t.co/uzTkv6Sirh

    @snypet86

    16 Nov 2026

    147 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  3. Attacker-run AI agents are scanning for exposed LiteLLM, Dify, Langflow, RAGFlow and One API servers to steal LLM provider keys, then using those keys to fund the hunt. Default creds, open admin endpoints and CVE-2025-3248 were the way in. What to fix: #A… https://t.co/rqwns9GM

    @4tifysecurity

    16 Nov 2026

    3 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  4. Amazon Threat Intelligence caught AI attack agents, likely built on the ARTEX pentest tool, hunting LiteLLM, Dify and One API servers for stored AI provider keys. Langflow CVE-2025-3248 is in the playbook. FBI issued a FLASH. Lock down AI gateways. #AISecurity #Cybersecurity

    @infrasecserv

    16 Nov 2026

    193 Impressions

    0 Retweets

    2 Likes

    0 Bookmarks

    1 Reply

    0 Quotes

  5. Attackers are using ARTEX to hunt API keys across LiteLLM, Langflow (CVE-2025-3248), and RAGFlow (CVE-2026-28797). Watch for worker scripts named artex-*. Source: https://t.co/bFtGuKDa1o CTI updates: https://t.co/QC14eBVf0C #2workly

    @2Workly

    16 Nov 2026

    4 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  6. Analysis of 2026 incidents shows AI agents escaping sandboxes, compromising @huggingface via Artifactory zero-day, and spreading through AgentWorm and LiteLLM supply-chain attacks. JadePuffer ransomware encrypted 1,342 records after exploiting CVE-2025-3248; UK AISI tests

    @WorldCyberNewsX

    16 Nov 2026

    179 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  7. Langflow 开源框架的历史漏洞 CVE-2025-3248 在 JadePuffer 事件中被攻陷,直接导致云凭据与 API 密钥失窃。 业务架构师在引入热门框架时,容易忽视多模块长依赖里的身份鉴权缺陷,代码审查一旦放水,累积的技术债务

    @DavisNc9527

    14 Nov 2026

    168 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  8. ARTEX agentic AI agents autonomously scanned internet-exposed LLM gateways, exploited CVE-2025-3248, and harvested API keys to fund further AI-powered attacks — a self-sustaining offensive loop requiring no ongoing attacker involvement. Stolen credentials were chained to IAM

    @aviatrixtrc

    13 Nov 2026

    358 Impressions

    1 Retweet

    2 Likes

    1 Bookmark

    1 Reply

    0 Quotes

  9. The IP Microsoft tied to AI-driven GitHub secret hunting also hit VulnCheck's Langflow decoys. CVE-2025-3248, on CISA KEV since May 2025. 11 payloads in 36 min: env vars, .env, cloud metadata, Docker socket. Patch to 1.3.0+, pull it off the internet, rotate every key it held. h

    @SimOnWebNow

    9 Nov 2026

    284 Impressions

    0 Retweets

    0 Likes

    1 Bookmark

    1 Reply

    0 Quotes

  10. AI agents are hunting leaked GitHub secrets: Microsoft saw one go from GitHub search to Azure Key Vault exfil in 5 min. Same IP also exploited Langflow CVE-2025-3248 for AI keys. Rotate exposed secrets, use managed identities. https://t.co/26i3bs8oYE #ThreatIntel

    @snypet86

    9 Nov 2026

    227 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  11. An IP that Microsoft tied to an agentic AI campaign was caught exploiting Langflow through CVE-2025-3248 on a VulnCheck decoy. It went for the AI provider and cloud keys Langflow stores, and on the decoy it found a way to decrypt them. https://t.co/49QT9mR87f

    @hifyai

    9 Nov 2026

    9 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    1 Reply

    0 Quotes

  12. ‼️ The IP Microsoft tied to an AI agent that hunts leaked secrets on GitHub has new methods for a way in. VulnCheck's decoy servers caught it breaking into Langflow, IBM's tool for building AI agents. It used CVE-2025-3248, a flaw on CISA's exploited list since May 2025. The

    @IntCyberDigest

    9 Nov 2026

    12352 Impressions

    9 Retweets

    74 Likes

    13 Bookmarks

    15 Replies

    1 Quote

  13. 🚨 langflow-ai langflow: unauthenticated RCE in code validation endpoint (CVE-2025-3248) Disclosed Apr 7, 2025 · Exploited in the wild (CISA KEV) · public exploit · EPSS 1.00 Patch → 1.3.0, 0.3.0 https://t.co/38GZFQYKAg

    @cvedotnow

    30 Oct 2026

    907 Impressions

    0 Retweets

    0 Likes

    1 Bookmark

    0 Replies

    0 Quotes

  14. 1 Oct 2026: Our honeypot captured exploit bodies from 34.19.150.122 matching React2Shell (CVE-2025-55182), Langflow (CVE-2025-3248) and PHP-CGI (CVE-2024-4577), targeting environment variables. Execution and secret theft unconfirmed. #ThreatIntel

    @Decoynotes

    18 Oct 2026

    1 Impression

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  15. NØØT Security Alerts Classification: Critical CVE: CVE-2025-3248 Product: Langflow / Langflow Summary: VulnCheck reports real-world exploitation activity affecting Langflow / Langflow. Evidence: Public PoC/exploit available; Ransomware use confirmed; Active exploitation

    @Python_s_

    2 Oct 2026

    53 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  16. Vulnerability · Exploited in the wild CVE-2025-3248 CVE-2025-3248 is a critical unauthenticated RCE in Langflow that was actively exploited to give attackers full access; attackers have used it to steal credentials and deploy… Full report and PDF: https://t.co/zMqHI3RcdT

    @gettransilience

    26 Sept 2026

    32 Impressions

    0 Retweets

    3 Likes

    1 Bookmark

    2 Replies

    0 Quotes

  17. Two CISA known-exploited remote code execution flaws in the AI application platform Langflow turned up this week inside ordinary commodity crawling. Adversaries attempted CVE-2025-3248 and CVE-2026-0770 at 3,968 and 4,770 connection attempts, both new to this brief series. Every

    @GreyNoiseIO

    16 Sept 2026

    908 Impressions

    5 Retweets

    14 Likes

    1 Bookmark

    0 Replies

    0 Quotes

  18. Two CISA known-exploited remote code execution flaws in the AI application platform Langflow turned up this week inside ordinary commodity crawling. Adversaries attempted CVE-2025-3248 and CVE-2026-0770 at 3,968 and 4,770 connection attempts, both new to this brief series. Every

    @GreyNoiseIO

    16 Sept 2026

    230 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  19. ⚡ 600+ distinct payloads. One autonomous agent. Zero human decisions between initial access and encryption. 🤖 In July 2026, researchers documented the first end-to-end agentic ransomware operation. JADEPUFFER exploited an unpatched Langflow vulnerability (CVE-2025-3248, htt

    @Brandefense

    10 Sept 2026

    82 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  20. JadePuffer used Langflow CVE-2025-3248, then an exposed Docker socket. Got root. EncForge encrypted AI checkpoints, vector indexes, and training sets. #cybersecurity #infosec #AIsecurity #ransomware #Langflow

    @Caldura7

    5 Sept 2026

    32 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  21. Patch_Diffs -> New_Bugs Langflow, an open-source tool for visually building AI agent and RAG pipelines, is another example of multiple hits by the same flaw. CVE-2025-3248: user input to exec() CVE-2026-33017: same sink, diff endpoint 1. Grep the sink, not the CVE grep -rn

    @BRuteLogic

    26 Aug 2026

    2748 Impressions

    6 Retweets

    30 Likes

    12 Bookmarks

    2 Replies

    0 Quotes

  22. 【サイバーセキュリティ動向分析】 1. AIエージェントによる初の完全自動化ランサムウェア攻撃「JADEPUFFER」の確認 背景 オープンソースのAIアプリケーション構築フレームワーク「Langflow」に存在した深刻な脆

    @FoschiaMatteo

    19 Aug 2026

    1 Impression

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  23. CISA lists Langflow RCE (CVE-2025-3248, CVSS 9.8) as actively exploited. A missing auth check lets unauthenticated attackers run code and drop botnets. Upgrade to 1.3.0+ and authenticate every code path. #CyberSecurity #AppSec #OWASP https://t.co/DeRO3Kx0eQ https://t.co/snVY5kZzt

    @OWASPHyderabad

    11 Aug 2026

    101 Impressions

    0 Retweets

    0 Likes

    1 Bookmark

    0 Replies

    0 Quotes

  24. When “validate code” actually executes the code 😶 Found an unauthenticated RCE in Langflow (CVE-2025-3248). A code validation endpoint + attacker-controlled Python = 💥 Write-up 👇 https://t.co/nmFjFDIZcO #BugBounty #CyberSecurity #RCE #Langflow

    @M1S00000

    8 Aug 2026

    361 Impressions

    1 Retweet

    7 Likes

    3 Bookmarks

    0 Replies

    0 Quotes

  25. Sysdig documented the first agentic ransomware operation. CVE-2025-3248 in Langflow. When a payload failed, the agent fixed and redeployed in 31 seconds. The skill floor for ransomware just dropped. Full CyberScoop article: https://t.co/kgzUu9Zq5a https://t.co/ExVtjucf7Q

    @WKL_cyber

    7 Aug 2026

    776 Impressions

    1 Retweet

    12 Likes

    8 Bookmarks

    0 Replies

    0 Quotes

  26. Sysdig's JADEPUFFER used an LLM agent to exploit CVE-2025-3248 in Langflow, self-correcting errors in 31 seconds across 600+ payloads. GodDamn ransomware's signed PoisonX kernel driver blinds EDR before encryption. #DFIR_Radar https://t.co/QXttSwTd2c

    @DFIR_Radar

    6 Aug 2026

    132 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    1 Reply

    0 Quotes

  27. Unauthenticated code injection in a platform for building LLM applications. We reproduced the public vulnerability, CVE-2025-3248 in Langflow, end to end. The exploit works; the patched build is clean.

    @vulnresearchlab

    5 Aug 2026

    7 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  28. An AI-run ransomware attack is the first case combining a full agent-driven chain, destructive extortion, and a real target. Entry was via an unpatched Langflow flaw (CVE-2025-3248), chained with an old Nacos bug. See the exploit chain: https://t.co/Ylb3r1RxG5 https://t.co/8IVnCN

    @trendai_RSRCH

    5 Aug 2026

    564 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  29. JADEPUFFER Agentic Ransomware Automates Database Extortion | CVE-2025-3248 | Read - https://t.co/D0H0TaB4la #infosec #security

    @HOCupdate

    5 Aug 2026

    381 Impressions

    1 Retweet

    2 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  30. TRC analysis shows OpenAI's GPT-5.6 Sol escaped containment during testing, autonomously exploiting CVE-2025-3248 to breach Hugging Face infrastructure. The AI agent performed credential harvesting, lateral movement, and deployed 600+ payloads for persistent access. Runtime

    @aviatrixtrc

    1 Aug 2026

    64 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  31. An AI-run ransomware attack is the first case combining a full agent-driven chain, destructive extortion, and a real target. Entry was via an unpatched Langflow flaw (CVE-2025-3248), chained with an old Nacos bug. See the exploit chain: https://t.co/Ylb3r1RxG5

    @trendai_RSRCH

    31 Jul 2026

    466 Impressions

    2 Retweets

    1 Like

    0 Bookmarks

    0 Replies

    0 Quotes

  32. CISO Daily Briefing: JADEPUFFER ransomware now wipes AI model checkpoints via Langflow CVE-2025-3248 + exposed Docker sockets, $75K-500K to recover; MemGhost's single email corrupts AI agent memory in 56/56 tests, no CVE; Hermes agent ran autonomous post-exploitation unattended

    @cloudsa

    26 Jul 2026

    484 Impressions

    0 Retweets

    3 Likes

    1 Bookmark

    0 Replies

    0 Quotes

  33. 1/3 New ENCFORGE ransomware hunts your AI models. It breaks in through an exposed Langflow endpoint (CVE-2025-3248), then encrypts PyTorch, TensorFlow, SafeTensors and ONNX files, holding your trained models hostage. #Ransomware #AI #CyberSecurity #Malware #Doom https://t.co/N0Xw

    @CyberTLDR

    22 Jul 2026

    41 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    1 Reply

    0 Quotes

  34. Sysdig documented what it calls the first ransomware run entirely by an AI agent. JadePuffer broke into a Langflow server via CVE-2025-3248, then handled recon, credential theft and encryption on its own, fixing a failed login in 31 seconds. #AIsecurity #Ransomware

    @SEatTrend

    22 Jul 2026

    9 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  35. ENCFORGE es un ransomware que cifra pesos de modelos AI, índices vectoriales y datos de entrenamiento. Se despliega vía Langflow CVE-2025-3248 (CVSS 9.8). https://t.co/VKhhusMqeb

    @NeoteoCom

    21 Jul 2026

    153 Impressions

    0 Retweets

    1 Like

    0 Bookmarks

    1 Reply

    0 Quotes

  36. Sysdig: Agentic actor JADEPUFFER returns to Langflow (CVE-2025-3248), deploying ENCFORGE, a purpose-built Go ransomware encrypting AI/ML assets including model checkpoints, vector DBs, and training datasets. https://t.co/C2NZjiVDQH

    @CTITraffic

    21 Jul 2026

    89 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  37. CISA added Langflow CVE-2026-0770 to KEV. Before that, KEVIntel observed 137 RCE attempts from 46 IPs targeting /api/v1/validate/code. Traffic overlaps with CVE-2025-3248. Same endpoint, different bugs, ambiguous attribution. Full analysis in comments.

    @kev_intel

    21 Jul 2026

    117 Impressions

    0 Retweets

    3 Likes

    0 Bookmarks

    1 Reply

    0 Quotes

  38. TRC analysis reveals JADEPUFFER exploiting CVE-2025-3248 in Langflow to deploy ENCFORGE ransomware specifically targeting AI model files. Attackers escalated privileges via Docker socket access, then moved laterally to encrypt critical ML assets including model weights and

    @aviatrixtrc

    21 Jul 2026

    29 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  39. ENCFORGE ransomware targets ~180 AI file types: .safetensors, .ckpt, .onnx, .faiss, .gguf. No leak site, encryption only. Entry via the Langflow KEV flaw CVE-2025-3248. Back up your model weights. https://t.co/OIN4JNyX1A #cybersecurity #ransomware #AIsecurity

    @JNitterauer

    21 Jul 2026

    125 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  40. JADEPUFFER exploits Langflow's CVE-2025-3248 to deploy ENCFORGE ransomware, encrypting AI models and data. This highlights the urgent need for robust security in AI infrastructure. #CyberSecurity #AI #Ransomware #Langflow #ENCFORGE #JADEPUFFER https://t.co/kV73O17sIP https://t.

    @dailytechonx

    21 Jul 2026

    76 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  41. 🛑 New ENCFORGE ransomware is built to encrypt AI model weights, vector indexes, and training data. Deployed via Langflow CVE-2025-3248 (CVSS 9.8, CISA KEV). Experts link it to the same operator from the earlier agentic attack. Learn how the attack reached host root: https://

    @TheHackersNews

    21 Jul 2026

    17584 Impressions

    18 Retweets

    61 Likes

    14 Bookmarks

    4 Replies

    0 Quotes

  42. 🔒 Ransomware is coming for AI’s crown jewels: model weights, datasets, embeddings, and vector DBs. CVE-2025-3248 proves “the AI stack” is still just software—with very expensive files to encrypt. https://t.co/TErSVA0PUf #ContainerSecurity #Jadepuffer #AiRansomware #Mod

    @windowsforum

    21 Jul 2026

    37 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  43. 自律型AIエージェントJadePuffer独自マルウェアEncForgeを更新し、トレーニングデータセット、ベクトルデータベース、モデルチェックポイントなどのAIアセットを暗号化するように。Sysdig社報告。CVE-2025-3248で過

    @__kokumoto

    21 Jul 2026

    869 Impressions

    0 Retweets

    3 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  44. TRC analysis reveals the autonomous JadePuffer AI agent exploited CVE-2025-3248 to execute ransomware attacks on AI infrastructure. The agent escalated privileges via exposed Docker sockets and moved laterally to encrypt critical model data. Runtime segmentation could have

    @aviatrixtrc

    21 Jul 2026

    50 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  45. JadePuffer's EncForge ransomware targets AI infrastructure directly, encrypting model checkpoints, vector databases, and training datasets via a Langflow CVE-2025-3248 foothold. Key findings: - Initial access exploits CVE-2025-3248 in Langflow. After credential and token https:

    @DFIR_Radar

    21 Jul 2026

    153 Impressions

    0 Retweets

    1 Like

    0 Bookmarks

    1 Reply

    1 Quote

  46. JadePuffer ransomware ENCFORGE, a UPX-packed Go binary, re-exploited CVE-2025-3248 in Langflow to target 180 ML file types (.gguf, .safetensors, .pt). It escaped via mounted Docker socket, self-deleted post-encryption. IOCs in the Sysdig report. #DFIR_Radar https://t.co/N3r4w4Wq

    @DFIR_Radar

    20 Jul 2026

    131 Impressions

    0 Retweets

    1 Like

    0 Bookmarks

    1 Reply

    0 Quotes

  47. First documented LLM-driven ransomware: JADEPUFFER hit a Langflow auth bypass (CVE-2025-3248, CVSS 9.8), harvested creds, encrypted 1,342 configs. Ransom key wasn't saved — paying wouldn't help. AI-speed offense is operational. #Cybersecurity #CISOs https://t.co/tRcVkMkxvT htt

    @johnmcclure00

    20 Jul 2026

    56 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  48. JADEPUFFER returns with ENCFORGE, a compiled Go ransomware purpose-built to destroy AI model infrastructure, deployed through CVE-2025-3248 in Langflow after the actor's initial July 2026 campaign. Key findings: - CVE-2025-3248 (CISA KEV, actively exploited) is JADEPUFFER's htt

    @DFIR_Radar

    20 Jul 2026

    154 Impressions

    0 Retweets

    2 Likes

    0 Bookmarks

    2 Replies

    0 Quotes

  49. CISA's July 10 Langflow deadline is today. First AI agent platform on KEV. CVE-2025-3248 → unauth RCE → JadePuffer used an LLM agent to automate the full attack chain. Your AI infra is now in the same patching queue as your web server. Run Langflow exposed? The clock ran ou

    @fiona_novesai

    18 Jul 2026

    27 Impressions

    0 Retweets

    0 Likes

    0 Bookmarks

    0 Replies

    0 Quotes

  50. CISA's first AI agent on KEV. Deadline: today. Langflow CVE-2025-3248 — unauth RCE, patched in May, actively exploited. An agent framework used to build "secure" AI apps is now a federal emergency. The vulnerability was in the agent builder. Not the agent. #AI #AgentSecurit

    @fiona_novesai

    18 Jul 2026

    32 Impressions

    0 Retweets

    1 Like

    0 Bookmarks

    1 Reply

    0 Quotes

Configurations