Documented, not implied.
Maestro's pitch is defensible evidence — so the platform itself has to be defensible. This page covers how we handle data, govern AI, and protect academic integrity, in the language your security review actually uses.
Where we stand, stated plainly.
We label what is certified, what is aligned, and what is underway — inflated claims fail procurement, and they should.
SOC 2
Controls are designed against the SOC 2 Trust Services Criteria; independent audit is on the roadmap and documentation is available under NDA.
GDPR
Data-protection by design: lawful-basis mapping, data-subject rights workflows, and EU data-residency options for European institutions.
FERPA
Student education records are handled under institutional control. Maestro operates as a school official with a legitimate educational interest.
Encryption
TLS 1.2+ for every connection; AES-256 at rest. Secrets are managed, rotated, and never present in application code.
Access & SSO
Learners and staff arrive through your identity provider. Role-based access, least privilege, and full administrative audit logs.
Accessibility
Accessibility is a design law, not a retrofit: every lesson follows one phone-first, accessible lesson standard.
Integrity by design, not surveillance.
The defense stack makes honesty the rational choice: tasks ask what only the learner knows, answers don't travel between learners, and behavioural signals route support — they never accuse.
Honesty is the cheapest path
Incentives are shaped so the easiest route through is the real one.
Tasks ask what only the learner knows
Not what a screenshot or a chatbot can hand back.
No two answers travel
Per-learner variation means copying gets you nowhere.
Signals support, never accuse
Behavioural evidence guides help — it is not surveillance.
Every milestone ends in a spoken defense
The learner talks through their own locked work — a conversation, not a trap.
AI-native means AI-governed.
The adaptive engine is powerful precisely because it operates inside hard boundaries. These are the four your review committee will ask about.
The human gate
No academic decision is made by AI alone. Content, assessment, and grading standards are approved by your subject-matter experts before anything reaches a learner.
Chosen models, approved prompts
Every AI model is chosen in Admin Center from a catalogue, and every prompt is versioned and approved in a prompt registry. Nothing is hard-coded — no silent capability changes.
Auditability of adaptation
Every adaptive decision traces to the evidence that produced it. Learners see it in plain words; institutions can audit it.
Learner data boundaries
Learner data trains no external foundation models. Institutional data stays institutional — contractually and architecturally.
Six roles, one release path.
Access follows six roles, and nothing reaches learners without a signature and a release.
- IT Admin
- Academic Lead
- Dean
- SME
- Registration
- Student
Named course staff
Every course has an owner, co-authors, and reviewers. Who can change what is set by role, not by habit.
Owner signs, Dean releases
The course owner signs the final report; the Dean releases a frozen version, which becomes live and immutable. There is no direct publish.
Rollback to a released version
If something needs undoing, a course rolls back only to an earlier released version — never to an unreviewed draft.
Learner records stay put
Released versions are frozen, so a learner's work and grades stay tied to the exact version they took — a later release never rewrites them.
One source of truth, clear boundaries.
Postgres-first architecture, standards-based LTI 1.3 and SSO integration, and a governed content pipeline — documented for your technical review.
Get the security packet.
Data-flow diagrams, subprocessor list, DPA template, and the AI-governance policy — everything your security and procurement review needs.