Security

Effective Date: 2026-05-29 Last Updated: 2026-07-17 2026-07-17: aligned this page's data-flow descriptions with the corrected 2026-07-16 privacy disclosures (voice processed by OpenAI/Deepgram with OpenAI's up-to-30-day abuse-monitoring window; camera/handwriting images read by Google OCR; transcript retention described as it operates today) and corrected the subprocessor-certification summary to defer to the per-vendor Subprocessor List; no change to security practices.

Trust comes first

Scout is built for children and their families. We take our responsibility to protect privacy and security seriously. This page describes — in plain English — how we think about security, what data we collect, how we protect it, and the controls available to you.

For full legal details, see our Privacy Policy and Terms of Service.

Our Privacy and Security Principles

Principle	What it means
Privacy-by-design	We build privacy into the product from the start. Your child's voice is processed in the moment by our speech providers (OpenAI for the conversation, Deepgram for reading transcription) and is never stored in our own cloud — OpenAI keeps it up to thirty (30) days for abuse monitoring, then deletes it. Verbatim transcripts stay on the family device (shared transiently with Anthropic for post-session analysis); our cloud stores summaries and metrics, not verbatim speech.
Data minimization	We only collect data necessary to operate Lantern. We do not collect last names, addresses, phone numbers, or social media accounts of children.
Parent control	Parents control everything. You can view, export, correct, delete, pause, or withdraw consent at any time through the Parent App.
Security-first engineering	Industry-standard encryption, restricted access, audit logging, and SOC 2 Type II certified subprocessors.
Transparency	We document what we do, what we use, and how we protect it — including this page and our subprocessor list.

What We Collect (Short Version)

The complete list is in our Privacy Policy. The summary:

From you (the parent): account email, billing info, your sworn affirmation that you're the parent.

About your child: first name, age, grade, optional context notes you choose to provide.

During sessions: voice audio (processed in the moment by OpenAI and Deepgram, never stored in our own cloud; OpenAI keeps it up to 30 days for abuse monitoring; a short reading recording stays on the device up to a day or two), camera and handwriting images (sent to Google for text recognition (OCR); camera copies stay on the device ≤7 days, handwriting copies until you request deletion), text transcripts (on the device; deleted on your request).

In the cloud: aggregated session summaries, reading scores, books read, AI-generated narratives.

We never collect: voiceprints, biometric identifiers, last names, home addresses (other than billing), phone numbers, school names, or social media accounts of children.

How We Protect Data

In transit

TLS 1.3 for all client-server communications
HTTPS-only for the consent app, Parent App, and Site
HMAC-signed webhooks for inbound integrations (KWS, payments)

At rest

AES-256 encryption for Supabase storage
Encrypted backups
Row-level security in our Postgres database (tenant isolation)

Access controls

Service-role keys for daemon access; per-user JWTs for Parent App access
Multi-factor authentication for Scout employee access to production
Audit logging of every production access
Restricted access: production credentials are limited to a small number of authorized Scout engineers
Access review at each policy version bump (quarterly cadence planned)

Vendor security

Subprocessor certifications (most SOC 2 Type II) are listed per vendor and reviewed at onboarding and annually
Every subprocessor is bound by terms that restrict use of the data we share — none is permitted to train AI models on it
Public subprocessor list: subprocessors.md

Application security

Secret scanning on every commit
Dependency vulnerability scanning (Dependabot / equivalent)
Code review required for all production changes

Children's Privacy Specifics

In addition to the above:

Verifiable parental consent (VPC) is obtained before any data collection from a child, using FTC-approved methods.
No targeted advertising to children. None.
No selling or sharing of children's data for cross-context behavioral advertising.
No use of children's data to train AI models.
Mandatory age-gating in account creation (parent-only sign-up).
COPPA Safe Harbor: not currently a member; may pursue ESRB Privacy Certified or kidSAFE Seal in the future.

Incident Response

If we identify a security incident affecting personal information:

Triage: within 24 hours of detection
Containment: as fast as is safely achievable
Notification to affected families: within 72 hours of confirmed impact, by email
Notification to regulators: as required by applicable law (state breach notification laws, FTC if COPPA-related)
Postmortem: publicly available redacted version, within 30 days

Parent Controls

In the Parent App, you can:

View every session your child has had with Lantern (full transcripts, summaries, books read)
Export all of your child's data (JSON + CSV bundle, emailed to you)
Correct anything that's wrong (first name, age, grade, context notes)
Delete specific items or your whole account
Pause data collection (reversible — Lantern still works but doesn't store new data)
Withdraw consent at any time (deletes everything per the retention schedule)
Toggle voice retention (research cohort families only)
Re-read the consent notice and policy version history

Independent Verification

We do not currently hold third-party certifications. Programs we may pursue:

COPPA Safe Harbor membership (TBD: ESRB Privacy Certified, kidSAFE, iKeepSafe, PRIVO, Aristotle, or TRUSTe)
SOC 2 Type II (for Scout — TBD; status: not yet pursued)
Penetration test (TBD: annual)
Independent privacy audit (TBD: pre-launch)

Subprocessor certifications are listed per vendor on the Subprocessor List.

How to Reach Us

General security questions: security@lanternlearning.io
Privacy concerns: privacy@lanternlearning.io
Vulnerability disclosure: email security@lanternlearning.io — we welcome responsible disclosure
Address: 65 Mission Trail Rd, Woodside, CA 94062

For a vulnerability disclosure, we commit to:

Acknowledging your report within two (2) business days
Investigating and providing a status update within ten (10) business days
Not pursuing legal action against good-faith researchers

Updates to this page

This page is updated as our security posture evolves. Material changes are announced via the Parent App and email.

Questions? Email privacy@lanternlearning.io. To file a complaint, contact the FTC at reportfraud.ftc.gov or your state attorney general.