Attributes
Toxicity, severe toxicity, insult, threat, identity attack, profanity, sexually explicit. Per-attribute score 0-1.
Advertisement
Multilingual
Supports 17+ languages. Uneven quality across languages, best in English.
Advertisement
Free tier + volume
Free up to 1 QPS. Higher tiers paid. Widely used in news comment moderation.
Known biases
False positives on AAVE, identity terms. Ongoing bias remediation. Audit periodically.