EverJoyBack to home

AI Usage Policy

EverJoy AI Usage Policy

Effective Date: July 1, 2026
Last Updated: July 2026
Version: 2026-07
Entity: EverJoy Memory LLC d/b/a EverJoy


1. Purpose and Scope

This AI Usage Policy ("Policy") describes how EverJoy uses artificial intelligence ("AI") in the Student Growth Cloud platform, including data flows, consent requirements, limitations, and human oversight. This Policy applies to:

This Policy supplements the Privacy Policy, Terms of Service, and Parent Consent Agreement.


2. AI Features Overview

EverJoy uses AI to enhance memory organization, growth tracking, and yearbook creation. AI features are optional and require separate parent consent (ai_processing, version 2026-07).

2.1 Feature Catalog

FeatureDescriptionAI ProviderConsent Required
Auto-ClassificationCategorize memories by type, subject, categoryOpenAI gpt-4o-miniYes
OCR & VisionExtract text from images and PDFsOpenAI gpt-4o-miniYes
SummariesGenerate memory summariesOpenAI gpt-4o-miniYes
Growth Dimension TaggingTag artifacts across 5 growth dimensionsOpenAI gpt-4o-miniYes
Growth InsightsNarrative insights and recommendationsOpenAI gpt-4o-miniYes
Yearbook RecommendationsSuggest content for Personalized YearbooksOpenAI gpt-4o-miniYes
Achievement DetectionIdentify proud moments and achievementsOpenAI gpt-4o-miniYes
Temporal Growth DetectionCompare memories over time (≥90 days apart)OpenAI gpt-4o-miniYes

2.2 Growth Dimensions

Partner capture and growth tracking use five dimensions:

  1. Academic Growth — Subject mastery, learning progress
  2. Communication Growth — Writing, speaking, expression
  3. Creative Growth — Art, music, creative projects
  4. Personal Growth — Character, resilience, social development
  5. Memories & Milestones — Significant events and achievements

3. Consent Requirements

3.1 Separate AI Consent

AI processing requires explicit ai_processing consent (version 2026-07), distinct from:

Parents may use core storage, timeline, and yearbook features without AI consent.

3.2 Consent Collection

Consent is collected via:

3.3 Re-Consent

When AI practices or policy versions change materially, parents must re-consent via /re-consent before AI processing resumes.

3.4 Consent Enforcement

AI processing is gated by:

Implementation note: Consent should be verified at AI job enqueue time (enqueueAiJob()). Production checklist item 9.5 tracks this enforcement.

3.5 Revocation

Parents may revoke AI consent through:

Revocation stops new AI processing. Previously generated AI metadata remains until parents delete associated memories.


4. Data Sent to AI Providers

4.1 OpenAI Integration

EverJoy uses OpenAI API (model: gpt-4o-mini) for all AI features.

Configuration:

4.2 Data Transmitted

FeatureData Sent to OpenAI
ClassificationTitle, description, OCR text (≤2000 chars), filename, MIME type; image URL for images
OCR/VisionSigned URL to memory file (15-minute TTL)
SummariesMemory content and metadata
Growth insightsChild first name, school year, signal excerpts, milestone text, timeline summaries, parent notes
Yearbook recsMemory summaries and metadata for current school year
Growth detectionMemory metadata for temporal comparison

4.3 Data NOT Sent

4.4 OpenAI Data Use

EverJoy configures API calls consistent with OpenAI's business terms. Per OpenAI API data usage policies, API data is not used to train models. EverJoy does not opt in to training data usage.

Parents should review OpenAI's Privacy Policy and API Terms periodically.


5. AI Processing Pipeline

5.1 Trigger Events

AI processing is triggered by:

5.2 Processing Flow

Upload/Capture
    │
    ▼
Consent Check (ai_processing)
    │
    ▼
Enqueue ai_processing_jobs
    │
    ▼
Worker: POST /api/ai/process
    │
    ├── Classification
    ├── OCR/Vision (if needed)
    ├── Summary generation
    ├── Growth dimension tagging
    └── Store results in memory record
    │
    ▼
UI: Poll /api/ai/jobs/[memoryId]

5.3 Fallback Behavior

If AI confidence is below threshold (0.7 for classification), rule-based fallback is applied. If AI is unavailable (cost cap, API error), processing is queued for retry or skipped with user notification.


6. Quality Assurance

6.1 Confidence Thresholds

ThresholdValueAction
Classification confidence0.7Below: rule-based fallback
Growth dimension confidence0.6Below: route to QA queue
OCR confidenceN/ARetry with vision if text extraction fails

6.2 Human Review

Artifacts with growth dimension confidence below 0.6 are routed to QA queue (qa_pending status). Admin/partner review required before assignment to student.

QA items unresolved for >48 hours trigger admin alerts.

6.3 Partner Capture QA

Capture Partners do not make AI-based identity decisions. Identity is determined solely via QR/barcode scan. AI is used only for content classification and growth tagging after identity is established.


7. Limitations and Disclaimers

7.1 AI Outputs Are Not Professional Advice

AI-generated content—including classifications, summaries, growth insights, and recommendations:

7.2 No Automated Decisions

EverJoy does not use AI to make automated decisions producing legal or similarly significant effects on children. AI outputs are suggestions for parent review.

7.3 No AI-Based Identity

EverJoy never uses AI, OCR, or handwriting recognition to identify students in the Partner Capture Network. Student identity is determined exclusively through Student ID label scan.


8. Cost Controls

8.1 Platform Cap

Monthly AI spend is capped at $200 platform-wide. When cap is reached:

8.2 Monitoring

Admin console (/admin/ai) displays:


9. Security

9.1 Data in Transit

9.2 Data Storage

9.3 Worker Security


10. Transparency

10.1 User Notification

Parents are informed about AI features through:

10.2 AI-Generated Content Labeling

AI-generated summaries and insights are distinguishable from parent-entered content in the UI. Parents can view original content alongside AI outputs.


11. Prohibited Uses

EverJoy AI will not be used to:


12. Changes to AI Features

Material changes to AI features or data practices require:

  1. Policy version update
  2. Parent notification
  3. Re-consent via /re-consent

Minor improvements (model updates within same provider, performance optimizations) may be deployed with updated Policy notice.


13. Contact

PurposeContact
AI/privacy inquiriesprivacy@myeverjoy.com
General supportsupport@myeverjoy.com


Schedule A: AI Model Specifications

ParameterValue
ProviderOpenAI
Modelgpt-4o-mini
API endpointapi.openai.com
Max tokens (classification)500
Max tokens (summary)300
Max tokens (insights)800
OCR text limit2,000 characters
Signed URL TTL15 minutes
Monthly cost cap$200
Retry attempts3
FallbackRule-based classification

Schedule B: AI Prompt Categories

Prompt TypeFileData Inputs
ClassificationclassifyMemory.tsTitle, description, OCR, filename
SummaryprocessMemoryPipeline.tsMemory content
Growth insightsgrowth/insights/prompts.tsChild name, signals, milestones, notes
Yearbook recsgrowth/yearbook/service.tsMemory summaries, school year
Growth detectiongrowth/summary.tsTemporal memory metadata

Schedule C: Prohibited AI Uses

EverJoy AI will NOT be used for:

  1. Student identification from handwriting or photos
  2. Admission or enrollment decisions
  3. Grading or academic evaluation
  4. Disciplinary action recommendations
  5. Medical or psychological diagnosis
  6. Predictive profiling for non-Service purposes
  7. Content generation without parent consent
  8. Training third-party models on user data

AI Usage Policy version 2026-07. Aligned with EverJoy AI implementation in yearbox/src/lib/ai/.

Back to home