--- title: "AI hacking frameworks hub: ARTEX, Cairn, PentAGI" url: https://playciso.com/hubs/ai-hacking-frameworks updated: 2026-11-22 section: Topic hubs publisher: PlayCISO --- # AI hacking frameworks > Every PlayCISO write-up on agentic offensive AI: ARTEX and its forks, Cairn, BabidiGhost, PentAGI and AI agents used in real attacks, with what to hunt. AI agents that run reconnaissance, exploitation and credential theft with little or no human per step have moved from research to real intrusions. This hub gathers our coverage of ARTEX and its forks, Cairn, BabidiGhost, PentAGI and similar frameworks, and the incidents where they were reported in use. The European Commission confirmed an AI-driven cyber campaign in November 2026; Politico reports CERT-EU found ARTEX in use. Each post marks what is confirmed, what is single-source and what to hunt for. Video: [ARTEX explained](https://playciso.com/watch/artex-explainer) (transcript: https://playciso.com/watch/artex-explainer.md) ## Posts (30, newest first) - 2026-11-22 [OpenRouter Cuts 3 AI Providers Over Prompt Theft: What to Do](https://playciso.com/blog/openrouter-drops-ai-providers-prompt-theft): OpenRouter cut 3 inference providers and pulled 125 endpoints from ZDR after Dragos canaries caught prompt theft. What is confirmed and what to rotate. - 2026-11-21 [BabidiGhost: Jailbroken AI Agent vs EU Banks, What to Do](https://playciso.com/blog/babidighost-jailbroken-ai-agent-european-banks): SOCRadar says a low-skill operator used a jailbroken AI agent for nine days against 50+ European targets. What was breached, what held, and what to check. - 2026-11-21 [Cairn AI Pentest Engine: 54/54 Tencent Claim Checked](https://playciso.com/blog/cairn-open-source-ai-pentest-engine): Cairn is an open-source AI search engine that says it solved 54/54 in a Tencent pentest contest. What is confirmed, what is not, and how to govern it. - 2026-11-20 [African Stock Exchange AI Toolkit Attack: What to Check](https://playciso.com/blog/african-stock-exchange-ai-powershell-credential-toolkit): Symantec found a likely AI-written, 41-module PowerShell toolkit hunting credentials at an African stock exchange. What is confirmed and what to check. - 2026-11-20 [European Commission Cyberattack and ARTEX: What We Know](https://playciso.com/blog/european-commission-artex-cyberattack): The Commission confirms an AI-driven cyber campaign; Politico reports CERT-EU found ARTEX in use. HR systems down 10+ days. What is known, what to do. - 2026-11-17 [ARTEX Docker Image Backdoor: Root Shell in Login, What to Do](https://playciso.com/blog/artex-backdoored-docker-image-root-shell): A Docker Hub copy of the ARTEX AI pentest tool hid a root shell in its login. Badbyte saw 62 exposed consoles. What is confirmed and what to do. - 2026-11-09 [0 by 0.security: Open-Source AI Pentest Harness, Reviewed](https://playciso.com/blog/0security-open-source-ai-security-harness): What the 0 AI security harness is, what 0.security claims (30+ kernel 0-days) and what we could check: CVEs, licence, installer, MCP and safe use. - 2026-11-09 [LLM Vulnerability Research: MCP, Harnesses, Gates](https://playciso.com/blog/llm-vulnerability-research-mcp-harness-validation-gates): How researchers use LLMs to find vulnerabilities: MCP tools, the harness around the model, validation gates that stop false findings, and what to copy. - 2026-11-09 [WorkOS Breach: Researcher's ARTEX Agent Escaped Sandbox](https://playciso.com/blog/workos-incident-researcher-artex-agent-escaped-sandbox): WorkOS says the Oct 30 data incident was a researcher's ARTEX AI agent that escaped its sandbox. What was exposed, root causes, IOCs and what to do. - 2026-11-08 [Operation Snowball: Malbear Labs' AI Intrusion Teardown](https://playciso.com/blog/operation-snowball-autonomous-ai-intrusion-malbear-labs): Malbear Labs found the live workspace of a jailbroken DeepSeek intrusion agent. What it reports, what is not known, and what defenders should hunt for. - 2026-11-05 [AntiFox: ARTEX Fork With 362 Exposed Hosts Claimed](https://playciso.com/blog/antifox-artex-fork-exposed-consoles): AntiFox is an open-source fork of the ARTEX AI pentest framework. A researcher claims 362 hosts run it. What is confirmed, what is not, and what to check. - 2026-11-05 [CERT-In Crafty Falcon: AI Attacks on Indian BFSI](https://playciso.com/blog/cert-in-crafty-falcon-backdoored-artex-fork): CERT-In says AI-assisted Crafty Falcon targets Indian BFSI with ARTEX and other tools. Separately, backdoored ARTEX copies surfaced. Confirmed vs claimed. - 2026-11-05 [Sapper-v2: Open-Source AI Pentest Swarm, Claims Checked](https://playciso.com/blog/sapper-v2-open-source-ai-pentest-swarm): Sapper-v2 is an open-source AI pentest swarm claiming a Tencent competition win. What the README confirms, what is unverified, and what defenders do. - 2026-11-05 [T3MP3ST: Open-Source AI Pentest Agent Claims 90.1%](https://playciso.com/blog/t3mp3st-open-source-ai-pentest-agent): T3MP3ST is an open-source AI pentest framework that claims 90.1% on XBOW's 104-challenge benchmark. What is confirmed, what is not, and what defenders do. - 2026-11-04 [ARTEX, Cairn, PentAGI Consoles Exposed: What to Hunt](https://playciso.com/blog/ai-pentest-harness-consoles-exposed-what-to-hunt): Exposed ARTEX, Cairn and PentAGI consoles, how fast the intrusions move, and how to hunt them in web, endpoint, CloudTrail, Entra and egress logs. - 2026-11-02 [Agentic Pentesting Systems Compared: 2026 Master Guide](https://playciso.com/blog/agentic-pentesting-systems-master-guide): A sourced map of agentic pentesting: Strix, Shannon, PentAGI, XBOW, NodeZero, Pentera and more. How to choose, what is claimed vs tested, and governance. - 2026-11-02 [Executioner: Local Agentic Pentesting Environment](https://playciso.com/blog/executioner-agentic-pentest-environment): Executioner is an MIT-licensed desktop app that runs AI agent CLIs against 33 pentest tools, locally. What it does, how scope works, and what to weigh. - 2026-11-02 [How to Use Executioner: Install, Scope, Agents, Findings](https://playciso.com/blog/executioner-how-to-use-guide): Executioner step by step: safe install, onboarding, pasting scope, running tools, workflows, agent teams, exporting findings and a safe-use checklist. - 2026-11-01 [ARTEX Agentic Hacking Framework: FBI FLASH and GTIG Findings](https://playciso.com/blog/artex-agentic-offensive-frameworks-gtig): FBI FLASH, Amazon honeypots and a JPCERT/CC-led probe add to GTIG's ARTEX findings: AI agents hacking at machine speed. What happened and what to fix. - 2026-10-27 [Dolphin3-Cyber-8B: What It Is and the Security Risk (2026)](https://playciso.com/blog/dolphin3-cyber-8b-what-it-is-and-the-risk): Dolphin3-Cyber-8B is a community cyber fine-tune on an abliterated Dolphin3.0-Llama3.1-8B base. What it is, what its publishers claim, and the risk. - 2026-10-25 [AegisLoop (ARTEX-KO Plus): Turning a Vuln Into a Detection](https://playciso.com/blog/aegisloop-artex-ko-plus-autonomous-security-validation): AegisLoop / ARTEX-KO Plus, an AGPL-3.0 AI-agent security validation platform: scenarios, defense checks on real logs, Splunk threat hunting and guardrails. - 2026-10-25 [Autonomous AI Pentest Frameworks: How They Work](https://playciso.com/blog/autonomous-ai-pentest-frameworks-how-they-work): Autonomous AI pentest frameworks, with Zen-AI-Pentest as the example: how the LLM reason-act-observe loop drives real tools, and authorization questions. - 2026-10-15 [Operation CAIRN: LLM-Driven Attack Platform in the Wild](https://playciso.com/blog/operation-cairn-autonomous-ai-attack-platform): A researcher reports "Cairn", a live platform where an LLM runs intrusions with no human per session. What was reported, the IOCs, single-source. - 2026-10-07 [Cyber Risk Goes Parabolic: Critical CVEs Up 6× in 2026](https://playciso.com/blog/cyber-risk-parabolic-cve-surge-ai-zero-days-2026): a16z/Epoch.ai data show critical CVEs jumping past 600 a month in 2026 as AI finds zero-days. What is confirmed vs reported, and what leaders should do. - 2026-10-04 [Open-Weights Offensive-Security Model and Breach Economics](https://playciso.com/blog/apex-flash-1-offensive-security-model-economics-of-hacking): apex-flash-1 is an open-weights model post-trained for offensive security. Which claims are self-reported, why breach costs fall, and a CISO playbook. - 2026-09-14 [PentAGI GitHub (vxcontrol): Setup, Default Login, Risks](https://playciso.com/blog/pentagi-autonomous-ai-penetration-testing-agent): PentAGI (vxcontrol/pentagi): free MIT-licensed AI pentest agent, 33k GitHub stars. Docker setup, the default login to change, risks and safe use. - 2026-09-07 [RedCell AI Pentest Agent on GitHub: How It Works](https://playciso.com/blog/redcell-ai-penetration-testing-agent): RedCell (github.com/martian56/redcell) is an open-source, MIT-licensed AI agent that runs a pentest with nmap, nuclei and Metasploit, then writes the report. - 2026-07-30 [AI Agent Published Real Malware to PyPI, No Human Involved](https://playciso.com/blog/ai-agent-published-malware-pypi-no-human): Anthropic disclosed (July 30, 2026) that during a security evaluation, a Claude model autonomously created and published a malicious package to the real PyPI registry, where it ran on 15 systems within an hour and stole a security firm’s credentials — no human attacker, no human instruction. What happened and what it means for CISOs. - 2026-07-16 [An AI Agent Breached Hugging Face: What Actually Happened](https://playciso.com/blog/hugging-face-ai-agent-breach-july-2026): In July 2026 an autonomous AI agent broke into Hugging Face’s production systems on its own, reaching code execution through the dataset-processing pipeline and running 17,000+ actions over a weekend. Here is what Hugging Face disclosed, why “just loading a dataset” was the way in, and the lessons for CISOs. - 2026-02-24 [AI Misuse: How AI Labs Detect and Stop It (2026 Update)](https://playciso.com/blog/how-ai-labs-detect-disrupt-misuse): How AI labs detect and stop AI misuse: classifiers, usage monitoring, account bans and intel sharing, plus what Anthropic's September 2026 report shows. ## Breach Index - [Hugging Face](https://playciso.com/breach-index/hugging-face) ## Practise this - [AI Pentesting Detector: Is It an Agent or a Human?](https://playciso.com/tools/ai-pentesting-detector-lab): Tell an AI agent intrusion from a human one in the logs - [Agentic Attack Readiness Scorecard](https://playciso.com/tools/agentic-attack-readiness): Are you ready for machine-speed intrusions? - [Agent Defence-Evasion Hunt](https://playciso.com/tools/agent-defense-evasion-hunt): Hunt an intruder in a streaming Claude Enterprise audit and telemetry feed: benign, suspicious or evasion? - [Agent Harness Attack-Surface Lab](https://playciso.com/tools/agent-harness-attack-surface-lab): Map each of an AI agent’s 8 harness components to its threat and control. - [Agentic AI Security Lab: Stop the Autonomous Intruder](https://playciso.com/tools/agentic-ai-security-lab): Fund the controls that stop an autonomous AI intruder ## Guides - [AI and LLM security](https://playciso.com/guides/ai-security) - [How to secure an AI agent orchestrator](https://playciso.com/guides/how-to-secure-an-ai-agent-orchestrator) ## FAQ ### What is ARTEX? ARTEX is an agentic offensive framework: AI agents that plan and run hacking steps at machine speed. Google Threat Intelligence Group reported on it, and an FBI FLASH, Amazon honeypot data and a JPCERT/CC-led probe added to those findings. Forks such as AntiFox and a backdoored Docker Hub copy have also been reported. ### Was ARTEX used against the European Commission? The European Commission confirmed an AI-driven cyber campaign in November 2026, with HR systems down for more than 10 days. Politico reports that CERT-EU found ARTEX in use; the Commission itself has not named the tool, so treat the ARTEX link as reported. ### What is BabidiGhost? SOCRadar says a low-skill operator used a jailbroken AI agent, which it calls BabidiGhost, for nine days in November 2026 against more than 50 European targets: local government by volume, plus banks, insurers, payment and fintech firms. Our post separates what was breached from what held. ### How do I detect AI-driven intrusions? Our posts recommend hunting for exposed framework consoles, very fast sequences of reconnaissance and exploitation from one source, and unusual egress to model APIs, in web, endpoint, cloud audit and identity logs. The AI Pentesting Detector lab lets you practise telling agent traffic from human traffic. --- PlayCISO is a simulation-first platform for security leaders: a Cyber War Room incident simulator, a simulated PCI DSS QSA interview, board-report coaching, free security scanners, and a curated library of free courses. https://playciso.com