The ManageBac+ AI features allow schools to manage end user access to large language models, or LLMs.

This technical overview explains the main principles that support the responsible use of AI in ManageBac+, including how data is handled, how prompts are processed, and how safeguarding measures are applied.

Overview

ManageBac+ and Faria Education Group have prioritised responsible use and safety in the introduction of generative artificial intelligence features.

• ManageBac+ does not share personally identifiable information with third-party LLM models.
• Third-party LLM models can only access data that has been explicitly selected or entered by end users.
• User prompts and AI responses are not reviewed to train LLM models.
• Third-party licensing agreements prevent user prompts from being used to train LLM models.
• Queries and responses are logged and are only available to trained specialists for service improvement purposes.

Key Concepts / What’s New

Data Protection

ManageBac+ AI is designed so that personally identifiable information is not shared with third-party LLM models. External models can only access content that has been explicitly selected or entered by the end user.

• Personally identifiable information is not shared with third-party LLM models.
• Only user-selected or user-entered content is accessible to third-party LLM models.
• Prompts and responses are not used to train LLM models.

Data Processing

ManageBac+ does not transfer personal or sensitive data across regions in the application or use of LLM models. The data retention policy for the service hosting the LLM model or models is approximately 30 days.

Safeguarding

All queries sent to LLM models are subject to filters that use AI reasoning capabilities to help protect users from inflammatory or harmful content.

If a user enters a prompt that would produce content in violation of these safeguards, ManageBac+ AI returns a message indicating that it cannot fulfil the request.

Only the end user who submitted the prompt receives this message. No additional alerts or notifications are sent to other end users or to ManageBac+ support personnel.

Because these safeguarding filters are also supported by an LLM model, their performance is subject to the normal variations associated with the technology.

Prompt Engineering

AI-enabled features are supplemented with additional content maintained by specialist teams to help produce quality output. ManageBac+ AI may adjust prompt engineering and associated parameters, such as temperature, in order to improve output over time.

The information passed to LLM models that affects their reasoning includes:

• User prompts entered directly by typing or clicking
• System prompts predetermined by prompt engineers
• Parameter values, such as temperature

Logs

All queries submitted by end users to LLM models, and the corresponding responses, are logged and are only available to trained specialists. They are reviewed only for the purpose of improving prompt engineering and the user experience.

Warnings & Important Notes

• Personally identifiable information is not shared with third-party LLM models.
• The data retention period for the service hosting the LLM model or models is approximately 30 days.
• Only the end user who submits a safeguarded prompt receives the refusal message.
• No additional alerts or notifications are sent to other end users or to ManageBac+ support personnel.
• Safeguarding performance may vary because it is supported by LLM technology.
• Queries and responses are only available to trained specialists and are reviewed only to improve prompt engineering and the user experience.