One prompt. Four policies. Different answers.

Pick a policy and a prompt, then check it. The decision comes back the way your team sees it in Vindex: the outcome, every check against its line, the policy line that decided, and what happens next.

This is a simulation with canned results; real decisions come from the API in about 100 ms.

Start your 14-day trial

Policy

Adult platformadult-platformChat assistantchat-assistantFamily appfamily-appMarketplacestock-imagery

Prompt

The mini-skirt oneKeyword filters flag it. Vindex reads what's actually asked for.Explicit, but legalTwo adults. Whether that's allowed is your policy's call.A named celebrityHarmless here, but a real, identifiable person.A copyrighted characterTag it, review it or block it: your line.“She's 18 but…”A stated adult age next to child cues. The trick doesn't lower the score.A minor in a sexual contextShown as a description only. Blocked on every policy, on every plan.An adult, presented youngA stated adult in a school costume. Review or block is a policy setting.A Japanese promptScreened as written, with no translation step.A jailbreak attemptMade for the Chat assistant policy. Vindex scores the request, not the wrapper.Something benignMost prompts look like this, and they should be fast.Type your ownMatched to the nearest example above. Never sent anywhere.

Your own prompt

Pick a policy and a prompt, then press Check.

The decision comes back the way your team sees it in Vindex.

Ready when you are

See your own prompts scored, live, in twenty minutes.

Bring a hundred prompts you've argued about internally. Start the trial and run them yourself to see where your policy line actually sits.

Start your 14-day trial Read the docs

Vindex

Policy-based content screening. Your policy, enforced in about a hundred milliseconds.

All systems normal