{"id":"CMP-202611-hrios5","kind":"campaign","aliases":[],"related":["CVE-2026-28797"],"title":"Chinese-speaking attackers run ARTEX AI agents to steal AI model API keys and cloud logins from exposed LLM gateways and AI app builders (LiteLLM, One API, Dify, Langflow, RAGFlow, Open WebUI, Ollama) (Amazon)","description":null,"summary":"Amazon Threat Intelligence (13 Nov 2026): from 3 Oct to 5 Nov 2026 its MadPot decoys posing as LLM inference APIs caught ARTEX agents sending their own campaign memory. Chinese-nexus groups (moderate confidence) ran ARTEX agents on near-autopilot to find internet-exposed AI apps, pull the upstream provider API keys they store, test them and pool them; the stolen keys paid for the LLM inference driving further hunting. Cloud credentials were used to enumerate storage, managed AI model services, secrets and parameter stores. Targets: LLM gateways/proxies One API, New API, Veloera, Sub2API, LiteLLM; AI app builders Dify, FastGPT, Langflow, RAGFlow, Open WebUI, LobeChat, Ollama. Routes: unclaimed uninitialized installs, anonymous management and provider-configuration endpoints, default admin credentials (LiteLLM default master key), CVE-2025-3248 (Langflow), CVE-2026-28797 (RAGFlow), an isCliRequest header auth bypass in One API family systems, SSRF to cloud metadata, and secrets in client-side JavaScript, source maps, debug routes and env files. Keys were also harvested from GitHub repos and event archives, npm and PyPI tarballs, container image layers, notebooks, model-hosting spaces, paste sites and Telegram. Amazon notified affected AWS customers; which other organizations were hit is unknown, and outcomes the agents reported are not verified. A second campaign tried deepfake KYC bypass at two crypto exchanges (all recorded attempts rejected).","severity":"high","cvss":null,"epss":null,"exploited":true,"exploited_at":"2026-11-16T07:54:01+00:00","products":[{"name":"LiteLLM","fixed":[],"ecosystem":"pypi"},{"name":"One API","fixed":[]},{"name":"New API","fixed":[]},{"name":"Veloera","fixed":[]},{"name":"Sub2API","fixed":[]},{"name":"Dify","fixed":[]},{"name":"FastGPT","fixed":[]},{"name":"Langflow","fixed":[],"ecosystem":"pypi"},{"name":"RAGFlow","fixed":[]},{"name":"Open WebUI","fixed":[]},{"name":"LobeChat","fixed":[]},{"name":"Ollama","fixed":[]}],"ecosystems":[],"tags":["campaign","early-warning"],"not_affected_if":[],"how_to_check":"Do you run any of LiteLLM, One API, New API, Veloera, Sub2API, Dify, FastGPT, Langflow, RAGFlow, Open WebUI, LobeChat or Ollama reachable from the internet? Check for: setup/init pages still claimable; admin or provider-config endpoints answering without login; LiteLLM still on the default master key; Langflow unpatched for CVE-2025-3248; RAGFlow unpatched for CVE-2026-28797; One API-family requests carrying an isCliRequest header. Search access logs for the Campaign 1 IPs 104.28.225[.]223, 31.22.91[.]156, 150.136.38[.]110, 128.242.115[.]131, 216.176.239[.]231 (seen 14 Oct - 2 Nov 2026). Watch AI provider billing for unexpected spend or usage from unknown IPs. Look for provider keys in public repos, published npm/PyPI packages, container image layers, front-end JS and source maps.","fix":"Take AI gateways and AI app builders off the public internet or put them behind SSO/VPN; finish setup on any uninitialized install; change default admin credentials and the LiteLLM master key; patch Langflow (CVE-2025-3248) and RAGFlow (CVE-2026-28797); block SSRF to cloud metadata (enforce IMDSv2). If an instance was exposed, rotate every upstream provider API key and cloud credential it holds, then review provider usage and cloud logs (identity lookups, storage listing, secrets/parameter store reads, managed model calls). Block the listed IPs.","status":"open","published_at":"2026-11-13T00:00:00+00:00","first_seen_at":"2026-11-16T07:54:01+00:00","triaged_at":"2026-11-16T07:54:01+00:00","updated_at":"2026-11-16T07:54:25+00:00","category":"ai-tooling","confidence":null,"triage":{"cause":{"at":"2026-11-16T07:58:33Z","path":"stolen-credential","entry":"unknown","reason":"The record lists many routes (unclaimed installs, anonymous endpoints, default credentials, CVEs, header bypass, SSRF, exposed secrets) without saying which one any affected victim used, while keys stolen from victims were then used by the attackers to pay for further hunting.","confidence":0.55}},"analyst_status":"done","analyst_note":null,"search":"'-2025':186B '-2026':190B '-202611':37A '-28797':191B '-3248':187B '13':61B '2026':63B,70B '3':65B '5':68B 'admin':179B 'affect':249B 'agent':8A,81B,95B,262B 'ai':7A,11A,23A,105B,139B,156B 'also':223B 'amazon':35A,58B,247B 'anonym':171B 'api':13A,28A,41A,43A,78B,111B,150B,152B,200B 'app':24A,106B,157B 'archiv':230B 'artex':6A,80B,94B 'attack':4A 'attempt':280B 'auth':196B 'autopilot':99B 'aw':250B 'builder':25A,158B 'bypass':197B,273B 'campaign':54B,85B,269B 'caught':79B 'chines':2A,88B 'chinese-nexus':87B 'chinese-speak':1A 'client':211B 'client-sid':210B 'cloud':16A,131B,205B 'cmp':36A 'confid':92B 'configur':176B 'contain':235B 'credenti':132B,180B 'crypto':276B 'custom':251B 'cve':185B,189B 'debug':216B 'decoy':73B 'deepfak':271B 'default':178B,182B 'difi':29A,159B 'dify':46A 'drive':128B 'earli':56B 'early-warn':55B 'endpoint':177B 'enumer':136B 'env':219B 'event':229B 'exchang':277B 'expos':19A,104B 'famili':201B 'fastgpt':47A,160B 'file':220B 'find':101B 'gateway':21A 'gateways/proxies':148B 'github':226B 'group':90B 'harvest':224B 'header':195B 'hit':256B 'host':241B 'hrios5':38A 'hunt':130B 'imag':236B 'infer':77B,127B 'instal':170B 'intellig':60B 'internet':103B 'internet-expos':102B 'isclirequest':194B 'javascript':213B 'key':14A,112B,122B,184B,221B 'kyc':272B 'langflow':30A,48A,161B,188B 'layer':237B 'litellm':26A,39A,155B,181B 'llm':20A,76B,126B,147B 'lobechat':52A,165B 'login':17A 'madpot':72B 'manag':138B,172B 'map':215B 'master':183B 'memori':86B 'metadata':206B 'model':12A,140B,240B 'model-host':239B 'moder':91B 'near':98B 'near-autopilot':97B 'new':42A,151B 'nexus':89B 'notebook':238B 'notifi':248B 'nov':62B,69B 'npm':231B 'oct':66B 'ollama':34A,53A,166B 'one':27A,40A,149B,199B 'open':32A,50A,163B 'organ':254B 'outcom':260B 'paid':123B 'paramet':144B 'past':243B 'pool':118B 'pose':74B 'provid':110B,175B 'provider-configur':174B 'pull':107B 'pypi':233B 'ragflow':31A,49A,162B,192B 'ran':93B 'record':279B 'reject':281B 'repo':227B 'report':263B 'rout':167B,217B 'run':5A 'second':268B 'secret':142B,208B 'send':82B 'servic':141B 'side':212B 'site':244B 'sourc':214B 'space':242B 'speak':3A 'ssrf':203B 'steal':10A 'stolen':121B 'storag':137B 'store':114B,145B 'sub2api':45A,154B 'system':202B 'tarbal':234B 'target':146B 'telegram':246B 'test':115B 'threat':59B 'tri':270B 'two':275B 'unclaim':168B 'uniniti':169B 'unknown':258B 'upstream':109B 'use':134B 'veloera':44A,153B 'verifi':266B 'warn':57B 'webui':33A,51A,164B","nicknames":["ARTEX AI key theft"],"mitigations":[],"who_affected":null,"tier":"A","verified":{},"analyst_revisit_at":null,"fix_status":null,"audience":null,"plain_action":null,"pkeys":["api","artexaikeytheft","dify","fastgpt","langflow","litellm","lobechat","newapi","ollama","oneapi","openwebui","ragflow","sub2api","veloera","webui"],"ptokens":["api","dify","fastgpt","langflow","litellm","lobechat","new","ollama","one","open","ragflow","sub2api","veloera","webui"],"labels":["aitooling","campaign","earlywarning","high"],"campaign_type":"exploitation","search_a":"'-202611':37A 'agent':8A 'ai':7A,11A,23A 'amazon':35A 'api':13A,28A,41A,43A 'app':24A 'artex':6A 'attack':4A 'builder':25A 'chines':2A 'chinese-speak':1A 'cloud':16A 'cmp':36A 'difi':29A 'dify':46A 'expos':19A 'fastgpt':47A 'gateway':21A 'hrios5':38A 'key':14A 'langflow':30A,48A 'litellm':26A,39A 'llm':20A 'lobechat':52A 'login':17A 'model':12A 'new':42A 'ollama':34A,53A 'one':27A,40A 'open':32A,50A 'ragflow':31A,49A 'run':5A 'speak':3A 'steal':10A 'sub2api':45A 'veloera':44A 'webui':33A,51A","search_b":"'-2025':186B '-2026':190B '-28797':191B '-3248':187B '13':61B '2026':63B,70B '3':65B '5':68B 'admin':179B 'affect':249B 'agent':81B,95B,262B 'ai':105B,139B,156B 'also':223B 'amazon':58B,247B 'anonym':171B 'api':78B,111B,150B,152B,200B 'app':106B,157B 'archiv':230B 'artex':80B,94B 'attempt':280B 'auth':196B 'autopilot':99B 'aw':250B 'builder':158B 'bypass':197B,273B 'campaign':54B,85B,269B 'caught':79B 'chines':88B 'chinese-nexus':87B 'client':211B 'client-sid':210B 'cloud':131B,205B 'confid':92B 'configur':176B 'contain':235B 'credenti':132B,180B 'crypto':276B 'custom':251B 'cve':185B,189B 'debug':216B 'decoy':73B 'deepfak':271B 'default':178B,182B 'difi':159B 'drive':128B 'earli':56B 'early-warn':55B 'endpoint':177B 'enumer':136B 'env':219B 'event':229B 'exchang':277B 'expos':104B 'famili':201B 'fastgpt':160B 'file':220B 'find':101B 'gateways/proxies':148B 'github':226B 'group':90B 'harvest':224B 'header':195B 'hit':256B 'host':241B 'hunt':130B 'imag':236B 'infer':77B,127B 'instal':170B 'intellig':60B 'internet':103B 'internet-expos':102B 'isclirequest':194B 'javascript':213B 'key':112B,122B,184B,221B 'kyc':272B 'langflow':161B,188B 'layer':237B 'litellm':155B,181B 'llm':76B,126B,147B 'lobechat':165B 'madpot':72B 'manag':138B,172B 'map':215B 'master':183B 'memori':86B 'metadata':206B 'model':140B,240B 'model-host':239B 'moder':91B 'near':98B 'near-autopilot':97B 'new':151B 'nexus':89B 'notebook':238B 'notifi':248B 'nov':62B,69B 'npm':231B 'oct':66B 'ollama':166B 'one':149B,199B 'open':163B 'organ':254B 'outcom':260B 'paid':123B 'paramet':144B 'past':243B 'pool':118B 'pose':74B 'provid':110B,175B 'provider-configur':174B 'pull':107B 'pypi':233B 'ragflow':162B,192B 'ran':93B 'record':279B 'reject':281B 'repo':227B 'report':263B 'rout':167B,217B 'second':268B 'secret':142B,208B 'send':82B 'servic':141B 'side':212B 'site':244B 'sourc':214B 'space':242B 'ssrf':203B 'stolen':121B 'storag':137B 'store':114B,145B 'sub2api':154B 'system':202B 'tarbal':234B 'target':146B 'telegram':246B 'test':115B 'threat':59B 'tri':270B 'two':275B 'unclaim':168B 'uniniti':169B 'unknown':258B 'upstream':109B 'use':134B 'veloera':153B 'verifi':266B 'warn':57B 'webui':164B","search_c":"","cause_entry":"unknown","cause_path":"stolen-credential","disclosures_checked_at":null,"first_public_at":"2026-11-13T00:00:00+00:00","kev":null,"cvss_vector":null,"cvss_by":null,"poc_count":0,"sources":[{"source":"analyst","external_id":"CMP-202611-hrios5","url":"https://aws.amazon.com/blogs/security/amazon-threat-intelligence-observes-threat-actors-running-cybercrime-campaigns-with-artex-ai-agents/","severity":null,"published_at":"2026-11-13T00:00:00+00:00","modified_at":null,"first_seen_at":"2026-11-16T07:54:01+00:00","fetched_at":"2026-11-16T07:54:01+00:00","note":null,"cites":["https://aws.amazon.com/blogs/security/amazon-threat-intelligence-observes-threat-actors-running-cybercrime-campaigns-with-artex-ai-agents/"]}],"documents":[{"id":168263,"url":"https://aws.amazon.com/blogs/security/amazon-threat-intelligence-observes-threat-actors-running-cybercrime-campaigns-with-artex-ai-agents/","title":"Amazon Threat Intelligence observes threat actors running cybercrime campaigns with ARTEX AI agents","kind":"blog","source":"analyst","publisher":null,"text":"Amazon MadPot (Oct 3 - Nov 5, 2026) captured ARTEX AI agent requests sent to decoy LLM inference APIs, carrying campaign memory. Campaign 1: large-scale theft of AI model API keys and cloud credentials from internet-exposed LLM gateways (One API, New API, Veloera, Sub2API, LiteLLM) and AI app builders (Dify, FastGPT, Langflow, RAGFlow, Open WebUI, LobeChat, Ollama), via uninitialized installs, anonymous management/provider-config endpoints, default admin creds (LiteLLM default master key), CVE-2025-3248 (Langflow), CVE-2026-28797 (RAGFlow), isCliRequest header auth bypass in One API family, SSRF to cloud metadata, secrets in client JS/source maps/debug routes/env files; also GitHub, npm/PyPI tarballs, container layers, notebooks, paste sites, Telegram. Stolen keys power further agents. Campaign 2: deepfake KYC bypass attempts at two crypto exchanges (vendor rejected all recorded attempts). Moderate confidence Chinese-nexus groups. IOCs Campaign 1: 104.28.225.223, 31.22.91.156, 150.136.38.110, 128.242.115.131, 216.176.239.231. Campaign 2: 212.107.28.51, 103.142.140.158, 104.28.251.98, 104.28.251.99, 104.28.219.97, 125.103.213.41. Amazon notified affected AWS customers; outcomes reported by agents not verified.","published_at":"2026-11-13T00:00:00+00:00","relation":"describes","confidence":null,"note":null},{"id":168246,"url":"https://x.com/TheHackersNews/status/2122232175529766932","title":"@TheHackersNews: 🚨 Amazon uncovers AI agents hunting API keys across LiteLLM, Dify, and other AI ","kind":"post","source":"x","publisher":null,"text":"🚨 Amazon uncovers AI agents hunting API keys across LiteLLM, Dify, and other AI apps.\n\nCaptured logs reveal the playbook: exploit weak spots → steal API keys → use those keys to power more attacks.\n\nThe campaign's reported results remain unverified.\n\nFull story: https://t.co/y0RG68SeLI\n\nlikes 126 reposts 26","published_at":"2026-11-16T07:35:54+00:00","relation":"describes","confidence":null,"note":null},{"id":168245,"url":"https://x.com/TheHackersNews/status/2122232821351890997","title":"@TheHackersNews: One exposed AI gateway. Multiple provider API keys at risk.\n\nAmazon's captured l","kind":"post","source":"x","publisher":null,"text":"One exposed AI gateway. Multiple provider API keys at risk.\n\nAmazon's captured logs reveal AI agents probing LiteLLM, Dify, One API and others for default credentials, auth bypasses and known flaws. https://t.co/Ul49j9Pu4o\n\nlikes 6 reposts 2","published_at":"2026-11-16T07:38:28+00:00","relation":"describes","confidence":null,"note":null},{"id":168244,"url":"https://x.com/TheHackersNews/status/2122232893032579394","title":"@TheHackersNews: The scary part? These attacks don't need new exploits.\n\nAmazon found AI agents t","kind":"post","source":"x","publisher":null,"text":"The scary part? These attacks don't need new exploits.\n\nAmazon found AI agents targeting old security failures at machine speed: default passwords, exposed secrets, unpatched software and open endpoints.\n\nWhat defenders need to check now: https://t.co/e9BCjbKcv0\n\nlikes 5 reposts 3","published_at":"2026-11-16T07:38:45+00:00","relation":"describes","confidence":null,"note":null}],"disclosures":[],"exploit_sources":[],"notice":"Free to read, with fair use of about 60 pages a minute. For higher limits or API access: sales@infrastack.ai"}