Release Notes: Content Moderation Settings for Bolt Student Assistants
Released 09/13/2024
Overview
We’ve introduced a robust content moderation tool for Bolt Student Assistants, which automatically flags conversations containing potentially inappropriate, harmful, or manipulative content. This tool works across all conversation channels—SMS, email, and live chat—whenever an Assistant handles the conversation. It monitors both text and image messages, enhancing safety by identifying content that is concerning or attempts to manipulate the Assistant. With flagging, you can apply customized responses and actions without human monitoring.
Details
Settings Location: Navigate to Conversation Settings > Bolt Assistants to find and configure your content moderation settings.
Automatic Flagging: Automatically flags Bolt Assistant conversations containing:
Sexual content
Hate speech
Harassment
Self-harm references
Violent content
Attempts to manipulate the Assistant
Instructions conflicting with the Assistant’s intended behavior
Cross-Channel Support: Works across all conversation channels where Bolt Assistants are active.
Text and Image Monitoring: Text and image messages are analyzed, expanding moderation beyond text-based communication.
Customizable Flags: For each flag type, you can configure:
Message: The text displayed to the external participant when the conversation is flagged for this category.
Type: The inbox formatting used for the flagged conversation.
Alert: The conversation preview is highlighted in red
Warning: The conversation preview is highlighted in yellow
Actions: Choose one or both of these automatic actions:
Disable Assistant: Turns off the assistant for the flagged conversation
Block Conversation: Prevents the external user from replying
Manual Flag Management: Conversation participants can manage flags under the "Manage" tab of the conversation panel to remove, change, or add flags.
Filtered Inbox: You can filter your conversation inbox by flag.
Conversation Rules: The new "moderation flag condition" in Conversation Rules allows for additional automation, such as assigning flagged conversations to specific users. Note: For this to work, you must also use the "Disable Assistant" action.
Benefits
This enhancement automates content moderation, safeguarding user interactions even without active human oversight. By flagging potentially harmful content and offering customizable responses, you can efficiently maintain a safe, respectful environment for all students.