Infaniti AI Inc.
Is it safe to journal with AI?
By Gavin Fan · Updated: 2026-11-01
Is it safe to put private entries into an AI journal?
AI journaling privacy depends on the product's data handling and what you choose to write. Check encryption, model processing, training, access, and deletion as separate questions. No app should be called safe solely because its page says encrypted.
This guide explains DiraX's stated practices. It is not an independent security audit or certification. The Privacy Policy is the governing document; Responsible AI explains product boundaries.
What should you check before choosing an AI journal?
| Question | What a useful answer should explain | DiraX's position |
|---|---|---|
| Where is text encrypted? | Transit, storage, and when processing needs readable text | Encrypted in transit and in the database; not end-to-end encrypted |
| Does an AI provider receive entries? | Which feature sends which material, and why | Relevant entry or message text is processed for comments, analysis, and chat; account email is not sent with that feature text |
| Is writing used for AI training? | Training versus ordinary inference, and any opt-in | DiraX does not use journals to train or fine-tune AI models |
| Can people access production systems? | Limits and specific support, security, or legal exceptions | Restricted production access; journals are not routinely read by the team |
| Can you leave? | Export, deletion, and any retention exceptions | Export and account deletion in Settings; retention details and exceptions in the policy |
How is journal data encrypted?
Encryption in transit and at rest
Encryption in transit protects data moving between services. Encryption at rest protects stored data. These protections do not mean the service can never process readable text. DiraX needs the relevant content to display your writing and run the AI features you request.
How this differs from end-to-end encryption
End-to-end encryption has a different access model. DiraX does not claim it. It also does not claim zero-knowledge storage, offline AI, or an external certification based only on using an infrastructure provider. If those are requirements for you, compare products whose architecture meets them.
How do AI providers process an entry?
When you ask for companion comments, analysis, or chat, the relevant entry or message is used to generate the result. Provider processing and model training are different uses. A no-training promise does not mean the text never leaves your device.
DiraX does not send your account email with the text used for those AI features. That is not a guarantee that journal text is anonymous: an entry can contain names or other identifying details that you typed yourself. The current provider and service information belongs in the Privacy Policy, so that list is not duplicated here.
Can staff read my journal?
The team does not routinely read journals. Production-system access is limited to specific purposes, such as a support request, a security issue, or a legal requirement. A promise of limited access is different from a technical claim that access is impossible.
If you contact support, describe the problem without including unnecessary entry text. Do not assume an AI reply has the same confidentiality rules or professional duties as a clinician or lawyer.
Can you export or delete your writing?
Settings provides export and account deletion. Check the policy for retention details, backups, legal exceptions, and the contact route for rights requests. Deletion should not be described as instant erasure of every backup unless that is what the actual policy and system provide.
Before a long-term commitment to any journal, locate its export controls and inspect the available format. Export capability does not automatically mean another product can import the result without conversion.
What information should you avoid entering?
A diary can record feelings and events without including passwords, financial account details, government identification numbers, or another person's confidential records. Minimize identifiers that the reflection does not need. Consider how much detail you are comfortable asking any online service to process.
This is a practical privacy choice, not a claim that DiraX detects or removes every sensitive identifier automatically.
Does Free or Premium change privacy?
No. Plan choice changes usage limits and feature access, not DiraX's no-training promise or the fact that it is not end-to-end encrypted. See Free and Premium limits for the actual product differences.
FAQ
Does DiraX use my diary to train AI?
No. DiraX does not use your journal to train or fine-tune AI models. AI features still process the relevant text to generate the response.
Are companion creators able to read my entries?
Publishing or choosing an AI character does not make your diary public or give the character's creator access to your entries. See how companions work.
Is this guide a security certification?
No. It describes DiraX's product and policy boundaries. It is not proof of an independent audit, regulatory certification, or protection against every possible threat.