Ai Guardrails
Adversarial LLM testing — 30 scenarios across jailbreak, prompt injection, PII leak, hallucination, tool exfil, cost control. Forensic tier; requires your endpoint URL + auth.
One of 122 modules in the GateTest scan suite. Catches the issue before it reaches code review, and on paid tiers opens a pull request with the fix already written.
Endpoint leaked system prompt under role-play jailbreak scenario 4
Why we catch it
Where deterministic scanning stops and reasoning starts. Used sparingly, not by default.
The Ai Guardrails module sits in this category alongside 6 related modules. Together they form one of the layers of a GateTest scan — checks fire in parallel, findings cluster by root cause, and on paid tiers the AI auto-fix loop reads each finding, writes the fix, validates against the scanner, and opens a PR.
How GateTest covers ai guardrails
- Forensic tier only. Forensic ($399) only — it is part of the deeper nuclear suite. Full and Scan + Fix do not run it. No additional configuration.
- Free CLI.
npm i -g @gatetest/cli && gatetest --module aiGuardrailsagainst any local repo. No paywall on the scanning itself. - AI auto-fix PR. Scan + Fix tier opens a pull request with the fix, a regression test, and a pair-review by a second AI reviewer. Forensic Scan tier adds per-finding diagnosis and cross-finding attack-chain correlation.
- Honest confidence rating. Findings come with high / medium / low confidence so noisy patterns don't block the gate. The confidence-calibrator trainer reads customer suppressions and tightens rules over time.
Scan your repo for ai guardrails
Free preview of the headline findings. Pay per scan, not per seat — one-time scans never auto-renew.
Frequently asked questions
What does the Ai Guardrails module catch?
Adversarial LLM testing — 30 scenarios across jailbreak, prompt injection, PII leak, hallucination, tool exfil, cost control. Forensic tier; requires your endpoint URL + auth. Example finding: Endpoint leaked system prompt under role-play jailbreak scenario 4
Does GateTest fix Ai Guardrails issues automatically?
Yes — on the Scan + Fix tier ($199) and Forensic Scan tier ($399), the fix engine reads the finding, writes the fix, validates against the scanner, writes a regression test, and opens a pull request for your review.
Which tiers include the Ai Guardrails module?
Forensic ($399) only — it is part of the deeper nuclear suite. Full and Scan + Fix do not run it.
Can I run the Ai Guardrails module from the CLI for free?
Yes — install with `npm i -g @gatetest/cli` and run `gatetest --module aiGuardrails`. The local CLI is free; paid tiers add AI auto-fix and the cross-finding correlation work.