Executive Summary

Student-facing generative AI systems introduce a new category of risk into educational environments.

Unlike traditional tools, these systems generate non-deterministic (same inputs - different outputs), real-time outputs that cannot be previewed, consistently monitored, or reliably reconstructed after the fact.

Current school governance structures were not designed for systems with these properties.

This is not a question of content; it’s a question of governance.

The structural gap

In every established area where schools introduce risk (e.g., transportation, facilities, athletics), three conditions are met before deployment:

  • A responsible party is clearly defined

  • Liability is understood and insured

  • Incidents can be reconstructed and investigated

Student-facing generative AI systems do not consistently meet these conditions.

What Makes This Category Different

Generative AI systems used by students:

  • Produce outputs that are non-deterministic (same input may yield different results)

  • Operate through conversational interaction, not fixed content

  • Cannot be fully observed in real time or reliably interrupted based on awareness of the interaction by supervising adults

  • Often lack complete, reconstructable interaction logs

  • Are deployed through software updates outside institutional review cycles

These properties create a structural mismatch with existing duty-of-care frameworks.

The Resulting Condition

When these systems are introduced without corresponding governance:

  • Instructional influence occurs outside supervised, accountable structures

  • Responsibility remains with schools and parents

  • Authority over the interaction is partially or fully delegated to the system

This creates a separation between responsibility and control. This resulting condition is true for student use. Teacher-facing AI is meaningfully different as the teacher is the accountable adult.

Principle

Responsibility must remain paired with authority.

If an institution is responsible for outcomes affecting children, it must retain the ability to:

  • See what occurred

  • Understand how it occurred

  • Intervene when necessary

  • Be accountable for the result

Non-Delegable Nature of Responsibility

Institutional responsibility for students is not reduced or eliminated by:
  • System complexity

  • Technological novelty

  • Or the perceived benefits of a given tool or technology

The introduction of new or advanced systems does not alter the requirement that responsibility remains clearly attached to accountable adult institutions.  

Difficulty of governance does not remove the obligation to govern.

What This Framework Does

This framework focuses solely on:

  • Accountability

  • Institutional responsibility

  • Governance alignment

This framework does not:

  • Regulate speech or content

  • Restrict access to information

Why This Matters Now

Generative AI is being deployed rapidly across school-issued devices and student accounts, often:

  • Without formal institutional review of specific use cases

  • Without insurer-confirmed coverage

  • Without reliable mechanisms for reconstruction of student interactions

This creates a condition where risk is introduced without the corresponding systems that typically manage it.

Bottom Line

When a system can influence a student through real-time, non-deterministic interaction, it must be governed as a source of institutional risk, not treated as a neutral tool.