Skip to main content

Security & Responsible Use

Pentest Swarm AI is an offensive-security tool. It exploits what it finds and proves it. That power comes with responsibility.

Authorized use only

Pentest Swarm AI is designed exclusively for authorized security testing, bug bounty programs, CTF competitions, and educational research. You must obtain explicit written permission from the target system's owner before running any scan against it.

Unauthorized access to computer systems is illegal under the Computer Fraud and Abuse Act (CFAA), the Computer Misuse Act, and equivalent laws worldwide. The authors and contributors of this project accept no liability for misuse, damage, or any illegal activity conducted with this tool.

By using this software, you agree that you are solely responsible for ensuring your use complies with all applicable laws and regulations. Do not use this tool against systems you do not own or have explicit authorization to test.

Scope enforcement — and why it's not a substitute for authorization​

Pentest Swarm AI enforces --scope as defence in depth: scope is checked at the tool layer and again at the executor, so an in-scope constraint is not bypassable by an agent. Cleanup is always registered before execution — SIGINT, crashes, and budget exhaustion all trigger reverse-order cleanup.

Those guardrails reduce accidents. They do not grant you permission. You are still responsible for pointing the swarm only at targets you're authorized to test, and for setting a correct scope.

The operational guardrails​

Alongside scope, three run-time controls keep autonomous runs in bounds:

  • --safe-mode blocks destructive command tokens (rm, DROP, kill, chmod, …) before execution — required by programs that forbid risky automated actions.
  • --budget <usd> caps per-run LLM spend and winds the run down gracefully when reached.
  • The killswitch (dashboard STOP button, or q / Ctrl-C in the terminal) ends a run cleanly at any time.

See Cost & Safety for how these combine to make an unattended run safe.

Safe places to practice​

Want to see the swarm work without touching anyone else's systems? Use the bundled labs (crapi, juiceshop, vampi, dvga) — intentionally vulnerable targets that spin up locally in Docker and tear down after. See the Quick Start.

Reporting a vulnerability in Pentest Swarm AI​

Found a security issue in the tool itself (not a finding it produced against a target)? Please report it privately — do not open a public GitHub issue. Open a GitHub Security Advisory or see the repo's SECURITY.md for details.