AI that helps you reflect, not one that learns from you
Daylogue uses AI to spot patterns and prompt deeper thinking. Here's exactly how that works.
How AI processing works
A step-by-step look at what happens when you use AI features.
Decrypt on device
Your encrypted entry is decrypted locally
Send to AWS Bedrock
Content sent securely over HTTPS
AI processes
Claude analyzes and generates response
Re-encrypt result
Response encrypted before storage
Important disclosure
During AI processing, your content briefly exists as readable text at our AI provider (AWS Bedrock). This is necessary for AI to work. It is not stored or logged, but we want you to know it happens.
AI and your privacy
Everything you need to know about how we use AI.
How AI features work
When you use conversational check-ins or pattern detection, your content is sent securely to AWS Bedrock for processing. It analyzes your text and generates summaries, narratives, and insights. AI-generated summaries are stored server-side to power features like your daily narrative and pattern detection. Your raw journal vault entries and encrypted notes remain end-to-end encrypted. Journal mode, quick check-ins, QuickPulse, vision boards, and connections all work without any AI conversation at all.
Not used for training
AWS Bedrock does not store, log, or use customer inputs to train or improve their models. Anthropic (the model provider) never sees your data directly. Bedrock isolates it within our AWS environment. For voice check-ins, ElevenLabs operates with PHI mode enabled. Zero data retention on their end.
You choose how much AI you use
Daylogue works however you want it to. Write freely in journal mode. Log a quick mood with sliders or a single emoji tap. Use the vision board, track connections, build habits. None of that involves AI conversation. When you do want it, conversational check-ins, pattern detection, and your daily narrative are there. You can also set your conversation style to "Private journal" to turn off conversational AI entirely. Nothing that makes Daylogue useful is locked behind AI interaction.
What runs in the background
Here is what happens automatically: some lightweight processing runs when you check in, like suggesting tags for your entries. This helps organize your journal over time. You never have to accept those suggestions, and features like your narrative and "Between the Lines" insights are always optional to read. Nothing is forced on you. If you skip it, it just sits there quietly.
The honest limitations
During AI processing, your content briefly exists as readable text in memory at AWS Bedrock. This is necessary for AI to work. The content is not stored, logged, or persisted. But it does exist in plaintext during processing. You should know that.
What we promise
AI has real privacy tradeoffs. Here is what we commit to.
AI for reflection
Patterns spotted. Insights generated. Then forgotten.
AI Provider
AWS Bedrock (Claude) + ElevenLabs (voice, PHI mode)
Input/Output Logging
Disabled. No prompts or responses logged
Data Persistence
Content processed in-memory, not stored
Transit Security
HTTPS (TLS 1.2+) for all AI requests
Training Data
Customer inputs never used for training
End-to-End Encryption
Not possible for AI requests (plaintext required)
Note: True end-to-end encryption for AI requests is not possible because AI models need to read your text to process it. Transit encryption (HTTPS) protects data in transit, but the AI provider must be able to see the plaintext during processing.