Azure Ai Contentsafety Ts

What it does

This skill integrates Azure AI Content Safety, enabling your agent to detect and filter potentially harmful or inappropriate content. It assesses text for categories like hate speech, sexually suggestive content, violence, self-harm, and more, providing scores indicating the likelihood of each category being present. The skill allows you to define thresholds for filtering content based on these safety scores.

When to use it

• Content Moderation: Automatically flag or block user-generated content (e.g., forum posts, chat messages) that violates community guidelines.
• Safe Chatbot Interactions: Ensure chatbot responses and user inputs remain appropriate and avoid generating offensive or harmful language.
• Protecting Vulnerable Users: Filter content in applications designed for children or individuals at risk of self-harm.
• Compliance Requirements: Meet regulatory requirements related to content safety and responsible AI practices.

Key capabilities

• Content Categorization: Detects hate speech, sexually suggestive content, violence, self-harm, threats, and more.
• Safety Scoring: Provides numerical scores for each category indicating the likelihood of harmful content.
• Customizable Thresholds: Allows defining specific thresholds to filter content based on safety scores.

Example prompts

• "Analyze this text for potentially unsafe content: 'This is a test sentence.'"
• "What's the hate speech score for this message: 'I disagree with their views.'"
• "Filter out any responses with a violence score above 0.7."

Tips & gotchas

• Requires an active Azure AI Content Safety resource and API key to function.
• The accuracy of content detection depends on the quality and context of the input text.