{"slug": "aws-cairn-ai-agents-found-a-monai-label-zero-day-in-11-minutes", "title": "AWS: Cairn AI Agents Found a MONAI Label Zero-Day in 11 Minutes", "summary": "Amazon Threat Intelligence reported on its Security Blog on November 20, 2026 that a China-linked operator left an unauthenticated Cairn autonomous pentest server exposed on the internet and pointed its AI agents at East Asian universities with the objective \"obtain shell access,\" with the agents discovering an unpatched path traversal flaw in the MONAI Label medical imaging annotation server in about 11 minutes. In a follow-on session, the agents chained the bug to root by planting a sitecustomize.py Python startup hook through a training endpoint that spawned Python as root, and MONAI published advisory GHSA-qc22-rmgc-vg94 with a fix merged but no release yet. AWS said the campaign targeted university IP addresses in East Asia outside mainland China for roughly 48 hours starting September 8, 2026, after Amazon Threat Intelligence spotted the exposed Cairn server on September 10, 2026, with prompts in Simplified Chinese and script timestamps in China Standard Time.", "body_md": "**Amazon Threat Intelligence says a China-linked operator left a public [Cairn](https://github.com/oritera/Cairn?ref=explainx) autonomous pentest server on the internet, pointed agents at East Asian universities with the goal \"obtain shell access,\" and watched them find an unpatched path traversal flaw in [MONAI Label](https://github.com/Project-MONAI/MONAILabel?ref=explainx) in about eleven minutes.** In a follow-on session, the agents chained that bug to **root** on a medical-imaging server by planting a Python startup hook. AWS published the story on [its Security Blog on November 20, 2026](https://aws.amazon.com/blogs/security/chinese-nexus-actors-cairn-ai-agents-found-a-zero-day-on-their-own/?ref=explainx).\n\nThis is not the same October [NorForge retail hacking report](https://explainx.ai/blog/hacker-ai-agents-5000-dollars-deere-sharkninja-cummins-breach-norforge-cairn-2026) we covered from Reuters. AWS brings **primary telemetry**, **MONAI Label specificity**, and maintainer **notification** details defenders can act on today.\n\n## TL;DR\n\n| Item | Detail (per AWS Nov 20 post) | \n|---|---|\n| Tool | **Cairn** open-source autonomous AI pentest framework | \n| Exposure | Operator left Cairn UI **without authentication** | \n| Victim app | **MONAI Label** medical imaging annotation server | \n| Bug | **Path traversal** on upload endpoints; no CVE at time of exploit | \n| Speed | **~11 minutes** from API mapping to arbitrary file write | \n| Root | **sitecustomize.py** plus training endpoint spawning Python as**root** | \n| Advisory | [GHSA-qc22-rmgc-vg94](https://github.com/Project-MONAI/MONAILabel/security/advisories/GHSA-qc22-rmgc-vg94?ref=explainx) ; fix merged,**no release yet** | \n\n## How AWS discovered the campaign\n\nOn **September 10, 2026**, Amazon Threat Intelligence spotted a server hosting Cairn **without login**. Because the instance was public, outsiders could read **project configs, actor prompts, and agent output**. Prompts were in **Simplified Chinese**; script timestamps used **China Standard Time**.\n\nThe visible campaign targeted **university IP addresses in East Asia outside mainland China** for roughly **48 hours** starting **September 8, 2026**. AWS published two **indicators of compromise** IPv4 addresses (partially redacted in the post) from that window.\n\nCairn's design is straightforward: you give a starting point and a plain-language objective, and **multiple agents plan, execute, and adapt** without the operator clicking each step. AWS stresses the same UX that helps defenders also helps attackers who can pay model bills.\n\n## When agents fail for hours, then win in minutes\n\nCircular scan frame with a green highlight spot suggesting automated reconnaissance\n\nAWS highlights two **failed** long runs to show persistence:\n\n### Hardened Redis at one university\n\nAgainst an **unauthenticated Redis** instance, agents spent about **16 hours** across **148 sequential actions**. They tried replication abuse, sandbox escape, config writes, memory corruption, and even **Log4Shell** attempts against a co-located service. They also used **SSRF** through a **Qdrant** instance to map Docker networks and sprayed credentials against PostgreSQL, SSH, and other services. **None succeeded.**\n\n### GPU server credential spraying\n\nAnother operation ran **6+ hours** on a GPU server. Agents harvested roughly **80 usernames** from public faculty pages, papers, and lab sites, then tried **30,000+ password combinations** against SSH, an admin dashboard, storage, VPN, and a campus portal. **No working login.**\n\nThose failures matter: they show agents **iterate like junior pentesters** with unlimited stamina. The breakthrough came on a **soft target**: an internet-reachable **MONAI Label** install with default-style exposure.\n\n## The eleven-minute MONAI Label zero-day\n\nGreen lens highlights a spot on a stylized medical imaging ribcage outline\n\nAWS says agents identified the framework, **mapped API endpoints**, and focused on an upload route with a **filename parameter**. Path traversal tests succeeded. AWS assesses with **high confidence** this was a **zero-day** because no fix, issue, or database entry existed yet for the project.\n\nThe [GitHub advisory](https://github.com/Project-MONAI/MONAILabel/security/advisories/GHSA-qc22-rmgc-vg94?ref=explainx) now tracks the flaw:\n\n- Endpoints **`PUT /datastore/image`** and**` PUT /datastore/label`** join a caller-controlled ID onto storage paths without keeping writes inside the datastore.\n- **`MONAI_LABEL_AUTH_ENABLE=False` by default** , so any network client that can reach the port may trigger writes.\n- Versions **0.8.5 and earlier** are affected.\n- Fix logic landed in [**pull request #1915**](https://github.com/Project-MONAI/MONAILabel/pull/1915?ref=explainx) , merged** October 19, 2026** , but**patched versions: None** on the advisory as of**November 18, 2026** .\n\n### Root shell in a later session\n\nEleven minutes bought **arbitrary file write**. Root came later when agents:\n\n1. Wrote a **reverse shell** into**`sitecustomize.py`** (auto-imported by Python).\n2. Called a **training API endpoint** to spawn a fresh Python process.\n3. Watched that process import the payload and connect back as **root** .\n\nAWS does **not** claim patient scans were exfiltrated. It does show **machine learning research infrastructure** can become a **code execution bridge** when auth and network controls are loose.\n\n## Human operator re-enters after the zero-day\n\nIn another session, AWS says a **human stepped in** after the agents found the bug. The operator supplied a **proxy tunnel** into a **second university network** where the same vulnerable software ran, suggesting manual reuse of the exploit to pivot. Agents then explored **additional hosts** through that tunnel.\n\nThat pattern matches how offensive teams may use agents for **breadth** and humans for **lateral movement** once a repeatable exploit exists.\n\n## Attribution and parallel Cairn abuse\n\nAWS attributes the campaign with **moderate confidence** to a **China-nexus** actor, citing language, timestamps, and university targeting. It does **not** map the activity to a named APT.\n\nThe post also cites **Gambit Security** research on a **financially motivated** actor using Cairn against **online retailers**, underscoring that Cairn adoption is **not limited to espionage**. Our earlier [Reuters-based Cairn retail story](https://explainx.ai/blog/hacker-ai-agents-5000-dollars-deere-sharkninja-cummins-breach-norforge-cairn-2026) covered different victims but the same framework family.\n\nFor a wider agent-misuse index, see explainx.ai's running [**felony-bench**](https://explainx.ai/felony-bench) tracker of agents affecting real third parties.\n\n## Actions AWS and maintainers took\n\nAWS says it **notified MONAI Label maintainers**, who published a fix path and GitHub advisory. Victim universities were **not AWS customers**, so AWS worked through a **government partner** to alert them.\n\nBehind the scenes, AWS highlights **honeypots and network-level restrictions** on scanning activity. That is infrastructure-level defense, not a substitute for patching research software you expose.\n\n## Defender playbook for ML research stacks\n\nShield with a hairline crack under a green security patch\n\n### Network and auth defaults\n\nMONAI Label is built for **trusted lab networks**, not open internet edges. Until a patched package ships:\n\n- Set **`MONAI_LABEL_AUTH_ENABLE=True`** and limit**ANNOTATOR** roles.\n- Block public ingress; require VPN or private links.\n- Run the service account with **minimal write permissions** outside the datastore.\n\n### Speed assumptions\n\nAWS argues SOC timelines tuned for **human attackers** lag **autonomous agents** that never pause. Pair high-confidence detections with **automated containment**, with humans reviewing outcomes instead of every single tool call.\n\nCoding agents that can hit ML admin APIs need outbound controls. [AgentBeam](https://agentbeam.com/?ref=explainx) can **block dangerous agent actions** before they reach exposed research endpoints.\n\n### Patch tracking\n\nWatch [GHSA-qc22-rmgc-vg94](https://github.com/Project-MONAI/MONAILabel/security/advisories/GHSA-qc22-rmgc-vg94?ref=explainx) and the MONAI Label release page. A merged fix on `main` is **not** the same as a version your orchestrator can pull.\n\n## How Cairn compares to other agent offense stories\n\n| Incident | What differed | \n|---|---|\n| [NorForge retail campaign (Oct)](https://explainx.ai/blog/hacker-ai-agents-5000-dollars-deere-sharkninja-cummins-breach-norforge-cairn-2026) | Journalist-sourced, e-commerce focus, mixed models | \n| **AWS MONAI case (Nov)** | Primary AWS blog, medical ML stack, measured **11-minute** zero-day window | \n| [OpenAI CI root misalignment report](https://explainx.ai/blog/openai-july-19-agents-peer-artifacts-ci-root-misalignment-report-2026) | Benign eval agents misusing peer credentials | \n\n## What we cannot verify from AWS alone\n\nAWS describes its reconstruction from a **public Cairn instance**. Independent media have not re-tested the MONAI exploit chain. Clinical impact (PHI exposure) is **not claimed**.\n\nTreat **root on one server** as a warning for **any default-open MONAI Label** on your perimeter, not proof every install falls the same way.\n\n## What is CVE-2026-133968?\n\nOutlets including [The Hacker News](https://thehackernews.com/2026/11/attackers-ai-agents-found-zero-day-in.html?ref=explainx) report that the MONAI Label flaw now carries **CVE-2026-133968** with a **9.8** critical CVSS score. The GitHub advisory timeline still shows **patched versions: None** while maintainers prepare a release. Treat the CVE as the tracking ID for scanners even if your package manager cannot upgrade yet.\n\n### Workarounds from the advisory\n\nUntil you can install a fixed release, MONAI Label's advisory recommends:\n\n- Enabling authentication and withholding **ANNOTATOR** from untrusted users.\n- Keeping the server off public networks.\n- Running the process as an **unprivileged** user with writes limited to the datastore tree.\n\nThose controls would not stop a determined insider, but they block the **anonymous network client** path AWS describes.\n\n## Why Redis and MONAI sat on the same campaign\n\nThe Redis stretch matters for defenders who think \"we patched Redis years ago.\" AWS documents **148** distinct attempts against a misconfigured instance. Agents treated failure as signal, not stop. Only when they found a **web-facing ML admin API** with permissive uploads did the timeline collapse to minutes.\n\nIf your org runs **research GPUs** and **legacy data stores** on overlapping subnets, segment them. Agents chaining **SSRF from Qdrant** to internal PostgreSQL shows how graph-like exploration crosses service boundaries humans forget to firewall.\n\n## Red-team versus real attacker economics\n\nCairn's README celebrates defender use in hackathons. AWS and Gambit show the **same UI** on a VPS with a stolen API key budget. Your red team using Cairn with authorization is legitimate; an exposed Cairn dashboard without auth is a **live operation journal** for threat hunters.\n\nDocument that difference in policy: pentest agents belong in **isolated VPCs**, not port 443 on a default cloud image.\n\n## Clinical and research data boundaries\n\nMONAI Label often sits near **imaging pipelines** that are not HIPAA production systems but still hold identifiable research scans. AWS does not claim imaging theft in this post. The lesson is architectural: **labeling servers are production-adjacent** and attract agents scanning university IP space.\n\nPair this incident with [ECHR lawyer sanctions for AI-hallucinated citations](https://explainx.ai/blog/echr-romanian-lawyer-barred-two-years-ai-hallucinated-citations-november-2026) as another reason research IT cannot treat ML tooling as \"lab only\" without network controls.", "url": "https://wpnews.pro/news/aws-cairn-ai-agents-found-a-monai-label-zero-day-in-11-minutes", "canonical_source": "https://explainx.ai/blog/amazon-cairn-ai-agents-monai-label-zero-day-root-11-minutes-november-2026?utm_source=rss&utm_medium=feed&utm_campaign=blog", "published_at": "2026-11-21 00:00:00+00:00", "updated_at": "2026-11-21 17:42:20.512906+00:00", "lang": "en", "topics": ["ai-agents", "ai-safety", "ai-policy"], "entities": ["Amazon Threat Intelligence", "AWS", "Cairn", "MONAI Label", "GHSA-qc22-rmgc-vg94", "Project-MONAI"], "also_reported_by": [], "takeaway": "Autonomous pentest agents mapped MONAI Label's API and reached arbitrary file write via a path traversal on an upload endpoint in about 11 minutes, then escalated to root through a sitecustomize.py hook spawned by a training endpoint running Python as root.", "recommendation": {"advice": "Treat internet-reachable ML annotation and training servers as soft targets: patch the MONAI Label path traversal (GHSA-qc22-rmgc-vg94) and stop training endpoints from spawning Python as root, since agents chained both to root in minutes.", "playbook": "agent-security", "playbook_url": "https://wpnews.pro/playbooks/agent-security"}, "alternates": {"html": "https://wpnews.pro/news/aws-cairn-ai-agents-found-a-monai-label-zero-day-in-11-minutes", "markdown": "https://wpnews.pro/news/aws-cairn-ai-agents-found-a-monai-label-zero-day-in-11-minutes.md", "text": "https://wpnews.pro/news/aws-cairn-ai-agents-found-a-monai-label-zero-day-in-11-minutes.txt", "jsonld": "https://wpnews.pro/news/aws-cairn-ai-agents-found-a-monai-label-zero-day-in-11-minutes.jsonld"}}