Field manual · AI for authorized security testing · Edition November 2026 · Updated 2026-11-18

Phase · November 2026

AI for internal network pentesting: the best model, and how to use it

The best AI for an internal network test is Kimi K3.1 Pro, because the useful step is tying enumeration you already have into a path.

DeepSeek V4.1 Flash reads that same long output for less, and both should see the scope file first.

Updated November 18, 2026Authorized work only
01Method

Execution, in the range the plan already named

NIST SP 800-115 treats the internal test as execution against the targets planning listed. Show the model what nmap already recorded, plus the range from the scope. Group hosts, then ports, then the odd service. HexStrike AI fits when a person still chooses the next check.

Rules of engagement

Only on systems you own, a signed scope, a CTF or a lab. A model that answers is not permission.

02With a model

Ask what to review next, inside the range

DeepSeek V4.1 Flash groups a large capture. Kimi K3.1 Pro says which hosts belong in one path. If SCOPE.md and the file disagree, the model should stop.

Prompt

Here is enumeration output from the internal range in SCOPE.md. Group the hosts, name the services worth a second look, and list the next checks inside the scope. Do not add addresses that are not in the file. Do not write exploit code.

Directory notes go to Active Directory. One application goes to web applications. Other grades are on the home matrix.