AI Disclosure

Last updated: March 30, 2026

1. How AI Is Used in Lagom

Lagom uses AI for the following features:

• Task extraction: when you explicitly trigger AI processing on a note, AI analyzes your text to identify and extract structured tasks — including task descriptions, priorities, due dates, and project categorisation.
• Contextual chat: an AI chat assistant that can answer questions about your notes and tasks, provide suggestions, help you organise your work, and save new content to your daily notes when you share information via chat or voice. Chat sessions are initiated by you.
• Semantic search: AI generates mathematical representations (embeddings) of your notes to enable meaning-based search — so you can find notes by concept, not just exact keywords.
• Daily briefs and insights: AI generates daily productivity summaries including overdue tasks, priorities, quick wins, and pattern observations.
• Organization suggestions: AI analyzes your projects and suggests merging, archiving, or reorganizing them for better structure.
• Voice transcription: when you use the voice capture feature, your audio recording is sent to OpenAI’s Whisper model for speech-to-text transcription. The transcribed text is then inserted into the chat input for your review before sending. Audio data is processed in real time and is not stored by Lagom after transcription.

2. Consent-Based AI Processing

AI processing in Lagom is driven by your use of the application. Lagom does not scan or process your content independently — AI is invoked as a direct consequence of actions you take:

• Saving a note automatically triggers task extraction and embedding generation as part of the save flow.
• Opening a chat sends your messages and relevant note context to the AI provider.
• Requesting a daily brief sends task metadata (names, due dates, projects) for AI summarization.
• Recording a voice note sends the audio recording to OpenAI Whisper for transcription. You must explicitly tap the microphone button to start and stop recording.

Notes that you have not saved or interacted with are never sent to AI providers. There is no background crawling, indexing, or processing of your content beyond the actions listed above.

3. What Data Is Sent to AI Providers

When you trigger AI processing, the data flow depends on your plan:

Free and Pro users

Your note plaintext is sent transiently via the Lagom server to the AI provider. The content passes through our server only to facilitate the API call — it is not stored, logged, or cached by our server during transit. Once the AI provider returns its response, the plaintext is discarded immediately. Only the structured output (extracted tasks, chat responses) is retained.

BYOK (Bring Your Own Key) users

When using BYOK, your API key is stored encrypted on our servers. The server decrypts your key to make AI requests on your behalf. Your key can be updated or deleted at any time in Settings.

When using BYOK, your note content is sent to the AI provider through our server using your API key. The same data processing rules apply — content is processed transiently and not retained by the AI provider.

In both cases, the following data may be sent to AI providers:

• For task extraction: the text content of the note you are processing.
• For chat: your chat messages and relevant context from your notes and tasks. When the chat saves content to your daily note, the AI also receives the existing note content to append correctly.
• For daily briefs: a summary of your current tasks, due dates, and project structure.
• For organisation suggestions: your project names and task counts.
• For voice transcription: the audio recording from your microphone. Audio is sent to OpenAI Whisper for speech-to-text conversion and is not stored by Lagom after transcription completes.

What is NOT sent to AI providers: your password, email address, account settings, API keys, or any authentication credentials. Only content you explicitly create (notes, tasks, chat messages) is processed by AI, and only when you initiate it.

4. Embedding Vectors and Semantic Search

When you save a note, Lagom generates an embedding vector — a mathematical representation of your note’s meaning. These vectors are:

• Generated by sending your note’s plaintext to OpenAI’s text-embedding-3-small model.
• Stored server-side in our database to enable semantic search.
• Not reversible to your original text — an embedding is a fixed-length numerical array (1536 dimensions) that captures meaning but cannot be decoded back into your words.
• Used exclusively for search — they are not shared with third parties or used for any other purpose.

Embeddings enable you to search by concept (e.g., searching “grocery shopping” can find a note about “pick up milk and eggs”) rather than requiring exact keyword matches.

5. Task Metadata

When AI extracts tasks from your notes, the resulting task metadata (task names, priorities, due dates, project assignments, and status) is stored on the Lagom server. This metadata is used by server-side features including:

• Daily briefs and productivity insights.
• Push notifications and email digests (e.g., due-today reminders).
• Organization suggestions.
• Future features such as calendar synchronization.

Task metadata is distinct from your full note content. Your notes are encrypted at rest, while task names and associated metadata are stored in a form accessible to the server so that these features can function.

6. AI Providers

Lagom currently uses the following AI providers:

• OpenAI — used by default for task extraction, chat, embedding generation, and voice transcription. Models used include GPT-4o-mini (for free-tier users) and other models available through BYOK. Embeddings always use OpenAI’s text-embedding-3-small model. Voice transcription uses OpenAI’s Whisper model.
  OpenAI API data usage policy
• Anthropic — available as an alternative provider when you bring your own API key (BYOK). Anthropic’s Claude models can be used for task extraction and chat.
  Anthropic privacy policy

We may add support for additional AI providers in the future. This disclosure will be updated accordingly.

7. Bring Your Own Key (BYOK)

Lagom supports “Bring Your Own Key” (BYOK), which allows you to use your own API keys from OpenAI or Anthropic. When you use BYOK:

• Your note content is sent to the AI provider through our server using your API key.
• Your data is processed under your own agreement with the AI provider, not ours.
• Your API key is encrypted with AES-256-GCM before being stored in our database.
• BYOK usage is never counted against any trial or usage quota.
• You can choose your preferred model (e.g., GPT-4o, Claude Sonnet) via the Settings page.

You can add, update, or remove your BYOK key at any time from the Settings page. When removed, the encrypted key is permanently deleted from our database.

8. Free Trial AI Usage

If you do not provide your own API key, Lagom provides a limited free trial of AI features:

• Free-tier users get up to 200 AI requests per month.
• Free-tier requests use OpenAI’s GPT-4o-mini model by default.
• Usage is tracked per user, per calendar month.
• When the monthly quota is exhausted, AI features become unavailable until the next month or until you add your own API key (BYOK) or upgrade to Pro.

Free-tier AI processing uses Lagom’s own API key with OpenAI. Your note content is sent transiently through the Lagom server to OpenAI and discarded immediately after processing. The same data policies apply as described in this disclosure.

9. AI Model Training

Your data is not used to train AI models.

Lagom uses AI providers through their API endpoints, not consumer-facing products. Both OpenAI and Anthropic have explicit policies stating that data submitted through their APIs is not used for model training by default:

• OpenAI: “OpenAI does not use data submitted by customers via our API to train OpenAI models or improve OpenAI’s service offering.” (Source)
• Anthropic: “We do not train our generative models on Customer Inputs or Outputs.” (Source)

We encourage you to review the AI providers’ current policies directly, as they may change over time.

10. Accuracy and Limitations

AI-generated content in Lagom — including extracted tasks, chat responses, daily briefs, and organisation suggestions — is provided “as is” and may contain errors. Specifically:

• Task extraction may miss tasks, create duplicate tasks, or incorrectly parse priorities and due dates.
• AI chat responses may be inaccurate, incomplete, or contextually inappropriate.
• Daily briefs and insights reflect AI interpretation and may not accurately represent your actual priorities.
• Project categorisation is inferred by AI and may not match your intended organisation.
• Voice transcription may contain errors, especially with background noise, accents, or mixed-language speech.

You should always review and verify AI-generated content before relying on it. Lagom provides tools to edit, correct, and override any AI-generated output. Your corrections are fed back into the AI system prompt to improve future extractions for your account.

11. Controlling AI Usage

You have control over how AI is used with your data:

• Consent-based processing: AI only processes your content when you explicitly trigger it. You can write and store notes without any AI involvement.
• BYOK: use your own API key so that AI requests are made with your key. You control your API key and can remove it at any time.
• Edit and correct: you can edit any AI-extracted task, change priorities, reassign projects, and your corrections improve future extractions.
• Delete data: you can delete individual notes, tasks, and chat conversations at any time. Deleted content is no longer sent to AI providers.
• Account deletion: deleting your account removes all your data, including any content that was previously processed by AI.

12. Changes to This Disclosure

We may update this AI Disclosure as we add new AI features, change providers, or modify how AI processes your data. We will notify you of material changes by posting the updated disclosure on this page with a new “Last updated” date. For significant changes (e.g., new AI providers or materially different data processing), we will notify you via email.

13. Contact Us

If you have questions about how AI is used in Lagom, or concerns about your data being processed by AI, please contact us at:

Lagom
[email protected]