AI is optional
Every project can provide an equal-status AI-off or provider-unavailable route.
Trust architecture · School control · Fail-closed live use
Trust is not a badge. It is a boundary the product keeps.
Responsible AI in schools cannot rest on a child remembering a policy or a teacher watching every chat. Own Mind makes permissions, custody, evidence limits, safeguarding routes and shutdown conditions part of the product architecture.
Project contract · Before the work opensSchool-controlled
- HHuman-only decisions declared
- AIPermitted machine roles declared
- OFFEqual AI-off route available
- DSLRestricted safeguarding path separated
The non-negotiables
The educational promise is enforced by product rails.
Own Mind begins from a simple rule: the system should be able to say no to itself before the school has to recover from a bad decision.
The machine gets a job, not the whole problem
Inputs, exclusions, approval points and a stopping condition are visible for each material AI contribution.
No automated child judgment
No system grade, authorship accusation, capability score, rank, personality profile or inferred trajectory.
The work is transparent
Assigned teachers see the complete ordinary-work record; students see the same content about themselves.
Safeguarding is separate
Safeguarding content is excluded from the ordinary teacher record and routed through the school’s restricted DSL process.
Access is first-class
No required home connection, public exposure or AI use determines whether a student can complete.
Need-to-know by design
The right adult sees the right thing for the right reason.
Each layer answers a different question. Keeping them separate prevents routine teaching access from quietly becoming safeguarding or administrative access.
| Layer | What it contains | Who uses it | What it explicitly excludes |
|---|---|---|---|
| Ordinary learning record | Work, decisions, revisions, material machine jobs and outcomes | Assigned student and teacher | Duration, engagement, behavioral trend |
| Safeguarding route | Restricted signpost and authorised handoff state | DSL or deputy DSL under school process | Routine teacher transcript access |
| School control plane | Identity, roles, providers, permissions, retention and incidents | Named school administrators, DPO and IT roles | Default access to student artifact content |
| Capability record | Human-authored, moderated, context-bound development statements | Student and authorised educational viewers | Automated inference, score, rank or comparison |
Mapped, inspectable
Every trust statement should point to a control, an owner and a gate.
Public assurances are kept at the evidence state they have actually earned. A mapping can be useful without pretending that an unsigned or untested control is complete.
What the system can and cannot do.
Permissions, retention boundaries, access scope, content routes and shutdown behavior are represented in the architecture.
What the institution must choose.
Providers, roles, project permissions, retention and live-use scope remain explicit school-controlled decisions.
What must be current before live use.
Required legal, safeguarding, data, provider and operational evidence gates activation for the exact capability and cohort.
Governance over generation
Keep the thinking with the child—and the authority with the school.
Trust Own Mind for what can be inspected: a bounded product architecture, an honest evidence ceiling and school-controlled live use.