Last updated: 7 May 2026 Effective date: 7 May 2026
This AI Disclosure explains the artificial intelligence used in journail.app (the "Service") in plain language. It is provided by Tomaž Pernovšek s.p. ("we", "us") to help you understand what AI does in Journail, what its limitations are, and what you should and should not rely on it for.
This disclosure is made in line with Article 50 of Regulation (EU) 2024/1689 (the EU AI Act), which requires that users be informed when they interact with an AI system. The relevant transparency obligations apply from 2 August 2026; we publish this disclosure now because we think you should know either way.
1. You are interacting with an AI
Whenever you use any conversational feature of Journail — including the Morning Brief, the Evening Debrief, the journal-generation step, the archive search, or the goals advisor — you are interacting with an artificial intelligence system, not a human being.
We display a visual AI label in the user interface wherever AI generates a response. If you are ever unsure whether something was written by AI or by a human, assume AI, and contact us at info@journail.app if you'd like clarification on a specific surface.
2. What the AI does in Journail
The AI is the engine behind the four conversational moments of the day:
2.1 Morning Brief
- Reads your active goals (yearly, monthly, weekly), today's calendar events, and your task list (from the built-in task manager and, if connected, Todoist).
- Holds a short guided conversation with you about your priorities.
- Proposes a focused plan with a small number of high-leverage tasks, deliberately avoiding micro-task vomiting.
- Sends accepted changes (re-scheduled tasks, new tasks, completions) back to Todoist if connected.
2.2 Focus & Flow
- Stays out of your way during the day.
- No push notifications, no interruptions.
2.3 Evening Debrief
- Asks you a few open questions about how the day went.
- Composes a daily journal entry from the conversation.
- The entry is yours; you can edit, accept, or discard it.
2.4 Archive
- Searches your past entries when you ask.
- Surfaces "On this day" reminders from prior entries.
- Counts how often a goal has been mentioned in your entries (a simple, deterministic search; not a model judgement).
In all of the above, the AI is a conversational and compositional partner. It suggests, drafts, asks. It never executes irreversible action without your confirmation. The decision is always yours.
3. Which AI providers we use
The AI is provided by third-party large language model providers acting as our data processors under Article 28 GDPR:
- OpenAI Ireland Limited — models in the GPT-4.1 family, including GPT-4.1 and GPT-4.1 Nano.
- Anthropic PBC — Claude models, used for selected reasoning tasks.
We may add or replace providers in the future. The current list is also published in our Privacy Policy under "Sub-processors", and we update it when it changes.
4. What we do — and do not — do with your content
4.1 We never train AI models on your content
We do not, and we contractually prohibit our AI providers from, using your inputs or the model's outputs to train, fine-tune, or evaluate any AI model. This is enforced through:
- OpenAI's API zero-retention or short-retention configuration combined with their no-training-on-API-data commitment.
- Anthropic's Commercial Terms, which exclude commercial API content from training by default.
- Our own product design, which never sends your content to public-facing or training endpoints.
4.2 What we send to providers per request
When you use an AI feature, we transmit only what is needed to generate that specific response. Typically:
- Your current message in the conversation.
- A relevant slice of context: active goals, today's tasks and calendar events, and (where useful for the feature) recent journal entries.
- A system prompt that defines the assistant's behaviour.
We do not send your email address, full name, payment information, or unrelated journal history.
4.3 What providers do with that data
Providers process the request to generate a response and then either delete it immediately (zero-retention configuration) or hold it transiently for up to 30 days for abuse monitoring and operational stability, after which it is deleted.
5. Limitations of the AI
We want to be honest about what large language models can and cannot do.
5.1 The AI is not a professional
AI output in Journail is not professional advice. It is not medical, psychological, legal, financial, tax, business, or coaching advice in the regulated sense. Treat it as a thoughtful conversational partner that can help you think — not as a substitute for qualified human judgement.
If you are in distress or experiencing a mental health crisis, please contact a qualified professional or your local emergency services. Helpful resources in Slovenia include:
- 112 — general emergency.
- 116 123 — Samarijan emotional support helpline (free, anonymous, 24/7).
- Klic v duševni stiski — 01 520 99 00 (17:00–07:00 daily).
5.2 The AI can be wrong
Large language models are probabilistic. They can:
- Fabricate facts ("hallucinate").
- Misremember details from your archive.
- Misunderstand instructions.
- Produce plausible-sounding but incorrect reasoning.
We add safeguards (system prompts, retrieval-augmentation, validation), but we cannot guarantee correctness. Verify important details before acting on them.
5.3 The AI does not "know" you the way a friend does
The AI does not have memory across sessions in the human sense. What feels like "memory" is the result of us showing the model a relevant slice of your archive and goals at request time. The model has no opinions about you outside that context. It is not your therapist; it is not your coach.
5.4 The AI is not always available
If a provider has an outage or rate-limiting incident, the conversational features may be temporarily unavailable. The Service will tell you if this happens. Your archive remains available regardless.
6. Decisions and human agency
6.1 No solely automated decisions
The AI does not make decisions producing legal effects or similarly significant effects about you, within the meaning of Article 22 GDPR. It is a suggestion and composition tool. You decide what to act on.
6.2 You can always opt out of an AI feature
You can choose not to use any conversational feature. The dashboard, the manual journal, the goals system, and the archive can be used without AI. The Service will continue to function as a structured planner and journal — it just won't generate or reflect.
6.3 You can correct or delete what the AI wrote
Generated journal entries can be edited or deleted like any other entry. The AI's contribution is a draft for you to own, change, or throw away.
7. AI risk classification under the EU AI Act
For transparency, we classify Journail's use of AI as limited-risk under the EU AI Act. The relevant obligations include:
- Article 50(1) — informing users that they are interacting with an AI system. Done via in-app labels and this disclosure.
- Article 50(2) — labelling AI-generated content where applicable. Generated journal entries are visibly distinguishable in-app from manual entries.
- Article 50(4) — providing this information clearly and distinguishably, at the latest at the time of the first interaction.
We do not deploy any AI system that would qualify as high-risk under Annex III of the EU AI Act. We do not use AI for biometric identification, emotion recognition for legal purposes, social scoring, or any other prohibited or high-risk use case listed in Articles 5–6.
8. AI-generated content labelling
Where the AI composes content for you — most notably the daily journal entry produced by the Evening Debrief — the entry is visibly tagged as AI-generated. You can edit it, and an edited entry retains a "human-edited" tag in metadata. If you publish or share an entry outside Journail (for example by exporting and posting it), labelling that content as AI-assisted is your responsibility.
9. Your rights regarding AI processing
You retain all of your GDPR rights when AI processes your content. Specifically:
- Access the data you have provided that has been processed by AI: Settings → Export data.
- Object to processing under legitimate interests (Art. 21 GDPR): write to info@journail.app.
- Withdraw consent for special-category content you have voluntarily included (Art. 7(3) GDPR): delete the relevant entries or your account.
- Lodge a complaint with the Slovenian Information Commissioner (Informacijski pooblaščenec, www.ip-rs.si) or your local supervisory authority.
For full GDPR rights and procedures, see our Privacy Policy and Your GDPR Rights page.
10. How we monitor AI quality
We continuously evaluate AI behaviour through:
- Internal evaluation prompts ("evals") that test for hallucination, refusal patterns, tone drift, and safety on synthetic data — never on user content.
- Community feedback channels (in-app thumbs-down, info@journail.app).
- Periodic review of any safety-related incidents reported to us.
If you encounter a problematic AI response — for example, factually wrong, harmful, distressing, or off-topic — please tell us. The thumbs-down button on any AI message sends us a sanitised report (the message ID and the rating; not the surrounding journal content) so we can investigate.
11. Changes to this disclosure
When we change AI providers, materially change what data is sent, or introduce a new AI feature, we will:
- Update this disclosure and the Privacy Policy.
- Notify users by email and in-app where the change is material.
- For consent-based AI features, ask for renewed consent where required.
12. Contact
For any question about how AI is used in Journail:
Tomaž Pernovšek s.p. Dolinškova ulica 10a, 1000 Ljubljana, Slovenia Email: info@journail.app
We answer AI questions like we answer any other question — directly, in plain language, and in good faith.