CriteriaBot

CriteriaBot is a content moderation API that evaluates any content against user-defined criteria written in plain English, returning true/false verdicts. It solves the problem of enforcing custom rules on user-generated content, such as toxicity, spam, prompt injection, brand voice, compliance, and more. The product targets developers, data teams, and operations teams in B2B companies, from startups to enterprises. It is positioned as a cost-effective alternative to large flagship models, using a panel of AI models that vote to achieve consensus, with options for custom model panels and personalization.

Key features

• Define criteria in plain English
• Group related criteria for reuse
• Community library of criteria
• Submit content via website or API
• Real-time or batch processing
• Panel of AI models votes on verdicts
• Bring your own API keys for custom panel
• Personalize verdicts with examples
• Wire results into pipelines
• Detect prompt injection
• Content moderation for toxicity and harassment
• Flag spam and abuse patterns
• Enforce brand voice guidelines
• Screen content for compliance
• Monitor brand and PR mentions
• Route and triage messages
• Label data for datasets