Skip to content
Get Started

Pre-built policies, each scored from 0 to 1

Each policy looks for one thing, such as toxicity, spam or swear words, and returns its own score, which can flag once it reaches its threshold.

Get Started FreeSee Pricing
  • Every score in the response
  • 8 on every plan, 18 on Starter and up
POST/v1/moderations
Sample data
Text
Arrived on time and works as described. Thank you.

TemplatebasicNo template named, so the basic check runs.

PolicyScore
  • Toxicity0.01
  • Hate speech0.00
  • Swear words0.00
  • Violence0.00
  • Self-harm0.00
  • Sexual content0.00
  • Illicit & regulated goods0.00
  • Spam & advertising0.02
SafeNot flaggedPublishedAllow
flagged
false
triggered
[]
recommendation
"allow"

Sample data

One catalog for text and images

Nineteen policies run today, each marked with what it reads; the three tagged Soon are in development.

Acme Corp./Pre-built policies
22 policies
  • Toxicity & hate
    • ToxicityText
    • Hate speechText
  • Profanity
    • Swear wordsText
  • Violence & self-harm
    • ViolenceTextImage
    • Self-harmTextImage
  • Sexual content
    • Sexual contentTextImage
  • Personal data
    • Personal dataText
  • Spam & risky links
    • Spam & advertisingText
    • Links & contact detailsTextImage
    • Duplicate contentTextImage
    • URL riskSoon
  • Illicit & regulated
    • Illicit & regulated goodsTextImage
  • Made with AI
    • Made with AIAdvancedExperimentalTextImage
  • Lookalike characters
    • Lookalike charactersText
  • Off-topic
    • Off-topicText
    • Meaningless textText
  • Image quality
    • Screenshot instead of photoImage
    • Low image qualityImage
    • Stock or third-party watermarkImage
  • Advanced
    • Audio transcriptionSoon
    • Image verificationSoon
  • Custom
    • AI instructionsTextImage

Observe a policy before it decides

In Observe mode a policy returns its score but never triggers or counts toward severity, and a content rule can still read it.

Acme Corp./Listings · Policies
Sample data
Made with AImade_with_aiAdvancedExperimental
Mode
Threshold0.90
EffectReview
Weight0.60
One submission, the same score in both modes
SafeNot flagged AI not detectedPublishedAllow
Made with AI0.92
flagged
false
triggered
[]
recommendation
"allow"

Sample data

A trigger flags, or sends to review

A trigger's effect sets the verdict, and the weighted score sets the recommendation unless a content rule decides first.

effect
Flag makes the result FLAGGED. Review does not flag. It turns a triage Allow into Review.
weight
From 0 to 1. The score is multiplied by it before triage.
Severity triage
Sample data
  • ToxicityFlag0.94 × 1.00 = 0.94FlaggedReject
  • Spam & advertisingReview0.75 × 0.80 = 0.60SafeReview
  • Low image qualityReview1.00 × 0.50 = 0.50SafeReview
  • Allow below 0.40
  • Review from 0.40
  • Reject from 0.80

Triage decides only when no content rule matched. A template can move 0.40 and 0.80.

Sample data

Eight policies run on every plan

On Starter and up your templates choose from 18 policies, and Made with AI makes 19 on Growth and up.

Compare plans
Which policies run

Basic check

Built in

basic runs when a request names no template

8 policies, fixed settings
ToxicityHate speechSwear wordsViolenceSelf-harmSexual contentIllicit & regulated goodsSpam & advertising

Seven flag at their default thresholds. Spam & advertising sends to review instead.

Your template

Starter and up

Runs when a request names it, or as the project's default

Can turn on those eight, plus
Personal dataLinks & contact detailsDuplicate contentLookalike charactersOff-topicMeaningless textAI instructionsScreenshot instead of photoLow image qualityStock or third-party watermarkMade with AIon Growth and up

Mode, threshold, effect and weight are yours to set, policy by policy.

Start moderating your users' content

Create an API key, send your first request and get a verdict back.

Get Started FreeSee Pricing