AI Safety Policy | Trivium Tutor

Trivium Tutor uses AI to teach children history and literature through Socratic dialogue. We've designed our AI with strict boundaries to ensure it remains a safe, focused educational tool. This document explains our AI safety principles and the guardrails we've built.

Core Principle: Education, Not Counseling

Our AI teaches history and philosophical ideas through texts and characters. It is not a therapist, counselor, or friend. All conversations focus exclusively on curriculum content - the stories, the people in them, and the ideas they represent.

This boundary exists to protect children and keep the learning experience focused and appropriate.

What Our AI Never Does

We've built explicit restrictions into our AI's behavior:

Never asks about personal experiences

The AI will not ask "Have you ever felt..." or "Tell me about a time when you..."

Never asks about feelings or emotions

The AI will not ask "How does that make you feel?" or "Does this remind you of your own life?"

Never asks about family, friends, or personal relationships

The AI will not ask "Who in your life is like this character?" or "What would your parents think?"

Never provides personal advice or opinions

The AI guides students to think, but does not tell them what to believe about non-academic matters.

What Our AI Does Instead

Focuses on the text and characters

"Why do you think the character made that choice?" instead of "What would you do?"

Discusses human nature in the abstract

"Why might someone act this way?" - inviting general reflection, not personal disclosure.

Redirects personal disclosures back to the text

If a student shares personal information, the AI acknowledges briefly and returns to the lesson content.

Uses Socratic questioning

Guides students to discover insights through questions rather than lecturing or telling them answers.

Example: How We Transform Questions

We would never ask:

"Have you ever felt left out like this character?"

Instead, we ask:

"Why do you think the character feels left out? What in the story tells us this?"

This keeps the focus on reading comprehension and critical thinking while still allowing children to develop empathy and understanding through the safe distance of fiction.

AI Data Privacy

Our AI interactions are designed with privacy in mind:

No personal data sent to AI: The AI receives only the lesson text and the student's responses about that text. It never receives email addresses, real names, or other identifying information.
Session-limited context: The AI only sees the current lesson conversation. It does not have access to other lessons, other students, or historical data.
No AI training on student data: Student responses are not used to train or fine-tune the AI model.
Parental visibility: Parents can review all AI conversations their children have through their account dashboard.

Age-Appropriate Content

Our curriculum is carefully selected and structured by grade level:

Grammar (Grades 3-4): Narrative histories written for children, focusing on stories and characters.
Dialectic (Grades 5-8): Primary source excerpts with guided analysis, building critical thinking skills.
Rhetoric (Grades 9-12): Complex texts requiring synthesis and argumentation.

All source texts are from the public domain (primarily Project Gutenberg) and have been selected for educational value and age-appropriateness.

Human Oversight

Curated curriculum: All lesson content and discussion questions are designed by humans. The AI guides conversation within these boundaries.
Parental access: Parents can view all lesson conversations and progress at any time.
Reporting: If you encounter any AI behavior that concerns you, please contact us immediately.

Questions or Concerns?

If you have questions about our AI safety practices, or if you observe any AI behavior that concerns you, please contact us immediately:

hello@triviumtutor.com

Related Policies

This AI Safety Policy works alongside our other policies: