AI SAFETY TESTTEXT PRE-SCREEN
Safety guideOpen app

Deterministic adversarial text pre-screen

Check risky text patterns before deeper AI testing.

AI Safety Test scans text you provide for configured adversarial signatures. It gives safety and product teams a clear starting point for human review.

Open the pre-screen runner
For

Safety, risk, and product teams preparing adversarial testing.

Use it to

Find configured phrases and nearby pattern combinations in supplied payload text.

Boundary

It does not call a target model, agent, tool, or API. A clean scan is not certification.

How it works

A visible pre-screen, then human judgment.

  1. 01

    Add payload text

    Paste the text you want to screen and choose a versioned set of adversarial patterns.

  2. 02

    Inspect matches

    See which configured phrases and proximity patterns appear in the supplied text.

  3. 03

    Record the review

    Document mitigation and the human decision before deeper model or agent testing.

Start with text you control

Run a focused pre-screen before the full safety evaluation.

Open AI Safety Test