Subprocessors

Third-party services that process Scout user data on our behalf. For each: what data flows, why, where it's processed, and the status of any contract (DPA / standard terms) we have with the provider. This page is the canonical transparency surface our privacy policy points to.

Last reviewed: 2026-05-16

DPA signedStandard termsDPA in progressNot applicable

Anthropic
United States
Large-language-model provider — generates Scout's Socratic responses to kid messages.
DPA in progress
Data flow
Conversation text (kid message + Scout system prompt + recent history). Per the Anthropic API terms, prompts and completions are not used to train Anthropic's models.
DPA note
Request workflow: Anthropic Console → Settings → Legal → request DPA. Update this row's status to 'signed' + the signed-on date once executed.
Last reviewed
2026-05-16
OpenAI (Whisper STT)
United States
Speech-to-text transcription for English voice sessions when voice mode is enabled.
Standard terms
Data flow
Audio clip of the child's spoken utterance. Returns the text transcript. Per OpenAI API terms, audio is not retained for model training.
DPA note
Standard OpenAI Business Associate / DPA terms apply via the API agreement. No separately-negotiated addendum.
Last reviewed
2026-05-16
ElevenLabs
United States
Text-to-speech synthesis for Scout's spoken voice responses.
DPA in progress
Data flow
Scout's response text (no kid input is sent). Returns synthesized audio.
DPA note
DPA request pending — outbound to ElevenLabs legal.
Last reviewed
2026-05-16
Microsoft Azure Cognitive Services
United States
Per-phoneme pronunciation scoring on language drill turns (Spanish / French / German vocabulary).
Standard terms
Data flow
Audio clip of the drill utterance + the reference text Scout is teaching. Returns word-level and phoneme-level accuracy scores. Only sent on drill turns — conversational turns are transcribed by OpenAI Whisper, never by Azure.
DPA note
Microsoft Online Services DPA covers Azure Cognitive Services with explicit COPPA and FERPA compliance posture.
Last reviewed
2026-05-16
Amazon Web Services (Lightsail + S3)
us-east-1 (United States)
Application hosting (Lightsail) and database backup replication via Litestream to S3.
Standard terms
Data flow
All Scout data lives on the hosting volume: user accounts, sessions, exchanges, voice clips, training audit rows. The SQLite database is continuously replicated to a private S3 bucket for disaster recovery.
DPA note
AWS Customer Agreement + DPA addendum apply to all Lightsail + S3 usage.
Last reviewed
2026-05-16
Cloudflare
Global edge (Cloudflare PoPs)
CDN + HTTPS termination for scout-learning.com (Flexible SSL mode currently; Full SSL upgrade tracked separately).
Standard terms
Data flow
Request headers and IP addresses pass through Cloudflare's edge for routing and DDoS protection. Cloudflare does not have access to application-level user data (kid messages, audio, etc.) — that traffic is decrypted only on the Lightsail origin.
DPA note
Cloudflare standard DPA applies via the account terms.
Last reviewed
2026-05-16
Resend
United States
Transactional email delivery — signup verification, magic-link sign-in, password reset, weekly digest, consent verification.
DPA in progress
Data flow
Recipient email address + email body (which contains links, no kid PII beyond the parent's name in some templates).
DPA note
DPA request pending — outbound to Resend.
Last reviewed
2026-05-16

How we maintain this list

Every third party with access to user data is listed here. If you find a service Scout uses that isn't on this page, please email [email protected] and we'll correct it.
Each row is re-reviewed at minimum during our bi-weekly audit cycle. The "Last reviewed" date at the top of this page is the most recent of all row reviews.
"DPA in progress" means we have requested a Data Processing Addendum from the provider and have not yet executed it. "Standard terms" means the provider's published terms include a DPA that applies automatically to our usage. "DPA signed" means we have a separately-executed agreement on file.
Adding a new subprocessor requires (a) updating this list before the integration ships and (b) updating the privacy policy if the data category is new. Both are tracked under the same PR.