Cybersecurity Research

AI-Powered Hacking Toolkits Are Spreading, and They Lower the Bar for Attackers

Cisco Talos · 14 Nov 2026
Key Takeaway Assume attacks will arrive faster and from less skilled actors, so patch internet-facing systems promptly and use multi-factor authentication to close the easy gaps these automated tools look for.

Cisco Talos is tracking the growing use of AI-enabled penetration testing and red team frameworks in the wild. These tools can automate and chain together reconnaissance, vulnerability testing and follow-on actions. When misused, they lower the skill needed to attack and speed up intrusions. Talos has seen full automation of the attack chain, using free and open-source tools like nmap and nuclei alongside internet mapping services such as Censys and Shodan, CAPTCHA bypass services, credential leak management services, residential proxies and even hash cracking services.

The frameworks come in different forms. Hephaestus is a command line tool that can be deployed with a single script on a VPS and controlled entirely through Telegram. PentAGI offers a web interface that shows agent activity as an attack chain graph, and Talos found more than 700 reachable PentAGI instances on the internet. HexStrike AI connects with clients such as Claude Code, VS Code and others that support MCP. ARTEX, one framework making waves, has let threat actors scale offensive operations, compressing work that once took skilled teams hours or days into activity that can unfold at machine speed. Talos also found a varied set of tools on ARTEX-associated IP addresses.

Talos expects these frameworks to evolve towards autonomous lateral movement, then data identification and exfiltration. It sees an early sign of this in the victim prioritisation of the Volta Infostealer. Talos considers it highly likely that future attack chains will end in data theft and ransomware activation, all carried out autonomously. This article covers only the opening of Talos's report, which goes on to discuss the challenges for defenders.

Summarised by CISO AI from Cisco Talos, written with Claude Sonnet 5.5. We link back to every original so you can read it yourself.