Information Security Policy

How we protect institutional data, for instructors and administrators reviewing Smart Learning System.

This page is generated from our source document. Institutional reviewers may request a signed or dated copy at admin@bayadataai.com.

Baya Data and AI LLC — Bothell, Washington, United States UBI 605 362 236

Policy owner Founder and Chief Executive (acting Information Security Officer)
Applies to All personnel, contractors, and systems used to operate Smart Learning System
Version 1.0
Effective September 22, 2026
Review cycle Annually, and upon any material change to the service or its subprocessors

1. PURPOSE AND SCOPE

This policy establishes how Baya Data and AI LLC ("Baya") protects the confidentiality, integrity, and availability of information entrusted to it, including student education records governed by the Family Educational Rights and Privacy Act (FERPA), institutional course materials, and Baya's own intellectual property.

It applies to all systems that store, process, or transmit institutional data, including cloud infrastructure, endpoint devices, source code repositories, and third-party services.

2. ROLES AND RESPONSIBILITY

The Founder serves as the accountable Information Security Officer and is responsible for the controls described in this policy, for reviewing it at least annually, and for approving any exception. Exceptions are recorded in writing with a justification and a target remediation date.

Where Baya has no employees, "personnel" means the Founder and any contractor granted access to institutional data. Contractors are granted access only under written confidentiality obligations no less protective than those Baya owes the institution.

3. DATA CLASSIFICATION

Class Examples Handling
Restricted Student education records, LTI user identifiers, assessment responses Encrypted in transit and at rest; access limited to the Founder; never exported to unmanaged devices
Confidential Institutional course materials, unpublished assessments and answer keys, source code, credentials Encrypted in transit and at rest; access limited to personnel with an operational need
Internal Operational documentation, architecture records Access limited to personnel
Public Website content, published certifications No restriction

4. ACCESS CONTROL

  1. Access is granted on a least-privilege basis and only for an operational need.
  2. Multi-factor authentication is required on all accounts with administrative or privileged access, including Microsoft Entra ID / Azure, Microsoft 365, GitHub, Stripe, and OpenAI.
  3. End users of Smart Learning System authenticate through their institution's identity provider by LTI 1.3 launch. Baya does not create or store passwords for LMS-launched users.
  4. Shared accounts are not permitted. Credentials are not transmitted by email or chat.
  5. Secrets and API keys are held in platform-managed secret storage, never in source code. Source control is scanned for committed secrets.
  6. Access is reviewed at least annually and revoked promptly when no longer needed.

5. ENCRYPTION

  1. All data in transit is protected with TLS. Plaintext protocols are not used.
  2. All data at rest is encrypted using platform-managed encryption provided by the cloud host.
  3. Endpoint devices used to access institutional data have full-disk encryption enabled.

6. NETWORK AND SYSTEM SECURITY

  1. The service runs on managed cloud services with vendor-maintained patching and platform-level network controls.
  2. Endpoint protection (anti-malware and host firewall) is enabled on devices used to operate the service.
  3. Administrative access to production is performed from managed endpoints over authenticated, encrypted channels. Baya operates no VPN or remote desktop services exposed to the public internet.
  4. Platform and dependency security updates are applied promptly; critical updates are applied without undue delay.

7. TENANT AND COURSE ISOLATION

Retrieval is scoped per course. A student's assistant draws only on material for courses in which the student is enrolled. Institutional data is logically separated so that one institution's content and student data are not retrievable by another.

8. SECURE DEVELOPMENT

  1. Changes reach production through version control and reviewed pull requests.
  2. Dependencies are monitored for known vulnerabilities and updated on a risk basis.
  3. Production credentials are never used in development or test environments.
  4. Institutional student data is not copied into development or test environments.

9. ARTIFICIAL INTELLIGENCE USE

  1. The service uses third-party large language model APIs to generate responses grounded in institutional course material.
  2. Institutional data submitted to model providers through the API is not used to train the provider's models. Baya does not fine-tune models on institutional student data.
  3. AI output is presented as assistive and non-authoritative. Instructors retain responsibility for academic and grading decisions, and may configure or disable assistant behaviour for their courses.
  4. Subprocessors used for inference are disclosed in the published Privacy Policy.

10. SECURITY AWARENESS TRAINING

All personnel complete security awareness training covering phishing and social engineering at least once every twelve (12) months, and upon engagement. The completion date is recorded by the Information Security Officer.

11. FUNDS TRANSFER VERIFICATION

To reduce the risk of business email compromise:

  1. Any request to create or change bank account details, payment instructions, or payee information is verified by an out-of-band callback to a previously known telephone number. A number supplied within the request itself is never used for verification.
  2. Any outbound payment at or above USD 1,000 to a new payee is verified by the same out-of-band method before processing.
  3. Verification is recorded, noting the date, the person contacted, and the number called.
  4. Urgency, secrecy, or authority claimed in a request does not waive these steps.

12. VENDOR AND SUBPROCESSOR MANAGEMENT

  1. Subprocessors that process institutional data are approved by the Information Security Officer before use and are listed in the published Privacy Policy.
  2. Licensed institutions receive advance notice of any new subprocessor that will process their student data.
  3. Subprocessor security posture is reviewed at onboarding and at least annually.

13. BACKUP AND RESILIENCE

  1. Institutional data is backed up using platform-managed backup services with point-in-time restore.
  2. Backup storage is encrypted and access to it requires multi-factor authentication.
  3. Restore capability is tested at least annually and the result recorded.

14. DATA RETENTION AND DISPOSAL

Institutional data is retained while the institution's agreement is in effect. Within sixty (60) days after termination, or upon the institution's written request, student data is deleted or returned, except where retention is required by law. Media and storage are disposed of using the cloud provider's secure deletion processes.

15. INCIDENT RESPONSE

Security incidents are handled under the Incident Response Plan, which defines severity levels, roles, containment steps, and notification commitments, including notification to affected institutions within seventy-two (72) hours of confirming a breach of their student data.

16. COMPLIANCE

  1. Baya acts as a school official with a legitimate educational interest under FERPA, 34 CFR § 99.31(a)(1), under the direct control of the institution with respect to the use and maintenance of education records.
  2. Baya holds 1EdTech Data Privacy and LTI Advantage certifications and maintains the practices those certifications require.
  3. Baya does not sell institutional or student data, and does not use it for advertising or for training third-party foundation models.

17. POLICY REVIEW

This policy is reviewed at least annually by the Information Security Officer, and on any material change to the service, its architecture, or its subprocessors. The review date and any changes are recorded in the version history below.


VERSION HISTORY

Version Date Change
1.0 September 22, 2026 Initial issue.