Researchers map 45 AI hacking frameworks and exposed consoles
Original title: Researchers Uncover 45 AI Hacking Frameworks, With Hundreds of Attack Consoles Exposed on the Internet
Shodan matches on November 7 included 1,428 ARTEX-family services; 11 of 41 fingerprinted tools expose a web UI by default.
Researchers catalogued about 45 open-source frameworks that loop a large language model through offensive tools. GitGuardian sorted 32 into five families, with 13 forks reusing those patterns. ARTEX’s creator removed the project on October 8, 2026, but forks continued, and published fingerprints matched many operator consoles online.
Unconfirmed: Reports that threat actors used ARTEX to hunt secrets and move laterally are second-hand.
- ARTEX-named repositories and forks numbered 102 in the takedown week.
- PentAGI showed 419 services, 233 using its default certificate.
- Agent Zero had 173 services, including 39 without a login.
- Cairn, HexStrike and Agent Zero ship with no authentication by default.
- Pojun, Cybermes, Xalgorix and AIRecon wire a real secret scanner to the agent.
Impact: Exposed agent consoles can run offensive tools and hunt secrets.
- 1.Check agent hosts for consoles on ports 8443, 8666, 8765, 9137 and 8787.
- 2.Enforce authentication on any operator UI.
- 3.Restrict outbound GitHub access from agent hosts.
- Organizations
- GitGuardian
From the original article
3 min read at gbhackers.com
Researchers have mapped about 45 open-source frameworks that wire a large language model into a loop that runs offensive tools and acts on their output. The catalog follows the takedown of ARTEX, an agentic penetration-testing platform. It shows the risk is spreading well beyond that single project.
By Mayura Kathir · gbhackers.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.