Umitake AI OCR for Kintone — Data Handling

The most important fact: documents submitted for OCR go directly to the AI provider you contract with. They never pass through an Umitake server, and we cannot see their contents at all.

The corollary is that you, not we, are the contracting party with the AI provider. Whether the data you send is retained, and whether it may be used for training, is governed by your agreement with that provider. We cannot influence it.

For the CDN and license verification common to every product, see the shared infrastructure document.

This document describes what is implemented, not what is intended.

1. Our Role

For this product we are not even a processor of your business data, because documents never reach our servers.

Traffic to the AI provider is relayed by kintone's own proxy feature (kintone.plugin.app.proxy), provided by Cybozu. Our servers are not on the path.

The AI provider is therefore your processor, not our subprocessor. Selecting, contracting with, and configuring that provider is something you do.

2. Data Flow

Your kintone
  └──> Your browser … file loading, PDF rasterisation
         └──> kintone proxy (Cybozu)
                └──> The AI provider you contract with
                       (Anthropic / OpenAI / Azure OpenAI / Google Gemini)
                         └──> OCR result → browser → saved into your kintone

Our servers appear nowhere in this flow.

3. What We Store

Nothing. We store none of it on our servers.

The only server we operate in this flow handles license verification, which receives three items: your kintone domain, the license key, and the plugin ID (see §3 of the shared infrastructure document).

Files awaiting OCR are held only in your browser's memory and are uploaded to your own kintone when the record is saved. Reloading the page discards them.

4. What Transits but Is Not Stored

Nothing transits our servers — but what is sent to the AI provider matters to you, so we set it out as implemented:

Easily overlooked: your master data is sent too

If you use the AI lookup search feature, records from the referenced app are included in the prompt for matching:

The fields sent are the key field and the lookup's display fields. In other words, your customer or product master list is transmitted to the AI provider on every OCR run that uses this feature. Review which fields you expose as display fields before enabling it.

Referenced records are cached in browser memory for 5 minutes.

5. Subprocessors and Locations

We have no subprocessors for this product. As explained above, the AI provider is your processor.

6. Retention Periods

We are not in a position to determine retention for OCR data.

Important: this product does not request zero data retention

The plugin sends no zero-data-retention header and no storage-disabling parameter to the AI provider. Retention and training use are governed entirely by your provider agreement and settings.

One point deserves particular attention: Google has published that the free tier of the Gemini API uses submitted content to improve its products. If you configure a free-tier API key, your documents will be used for training.

Enabling zero data retention or opting out of training is something you must arrange with your provider. If you handle business data, we strongly recommend a paid plan with the appropriate settings.

7. On Cancellation and Disconnection

There is no business data on our side to delete.

For data already sent to the AI provider, deletion must be arranged between you and that provider. We cannot act on your behalf.

OCR results are stored in your kintone, and we never touch them.

License records are handled as described in §7 of the shared infrastructure document.

8. What We Cannot See

• We cannot see the documents submitted for OCR. They never pass through us.
• We cannot see the OCR results. They are written straight into your kintone.
• We cannot see your prompts or any exchange with the AI provider.
• We cannot see your API key. You obtain it and it is stored in your kintone; it never reaches us.

9. Your Responsibilities

• Your agreement with the AI provider. You are the contracting party. Check that provider's terms on retention and training use, and configure zero data retention or training opt-out yourself (see §6).
• Managing the API key. It is stored in the kintone plugin configuration and is readable from the settings screen by anyone with kintone app administrator rights, so consider who holds that permission. Usage limits and billing for the key are yours.
• AI lookup display-field settings. As described in §4, records from the referenced app are sent to the AI provider. If you expose fields containing personal data, be aware that this data is transmitted too.
• Deciding which documents may be sent for OCR. Judge against your own information-handling policy whether documents containing sensitive information may be sent to an AI provider.

10. Contact

For questions about this document or help completing a security questionnaire:

Umitake Email: [email protected]

Related: Shared infrastructure · Privacy Policy · Terms of Service