Subprocessors
Version: 1.1 · Effective date: 2026-11-18
This page lists the third parties that process personal data when The Next Gen (جيل الغد), Erbil, Iraq, provides nashir.ai, the service it operates, to customers. It is Annex III of the nashir Data Processing Agreement.
How to read this list
- "Customer data" means data about a customer's own end customers: phone numbers, names, messages and media.
- A provider receives data only when the feature named in its section heading or its row is used.
- The processing locations and transfer mechanisms below are taken from each provider's public terms, read on 2026-11-12. The data processing terms of Anthropic, OpenAI, OpenRouter, Hostinger, Cloudflare and Resend apply automatically with their service terms; the agreement with Zoho was signed on 2026-11-12; Microsoft's addendum applies to Azure Speech without a separate signature.
1. Infrastructure (always used)
| Provider | Purpose | Data | Location | Transfer mechanism |
|---|---|---|---|---|
| Hostinger | The servers that run the application, nashir's database, file storage and nashir's automation engine; automated daily and weekly backups, stored separately from the servers | All customer data | Server: Germany. Backups: Lithuania. | None needed; data stays in the EU. Hostinger's group is based in the EU. |
| Cloudflare | Network proxy in front of the website, API and file storage; protection against attacks; secure channel for database administration | All traffic passes through it, including message content in API calls and stored media that it caches | Global network; requests are handled at the nearest location | Cloudflare's data processing terms with SCCs; Cloudflare is certified under the EU-US Data Privacy Framework |
2. Messaging platforms (when the customer connects the channel)
| Provider | Purpose | Data | Location | Transfer mechanism |
|---|---|---|---|---|
| Meta (WhatsApp Business Platform, Messenger, Instagram, Threads) | Delivering and receiving the customer's messages, comments, broadcasts and calls | Recipient phone numbers and user IDs, message content, media, templates | Meta's global infrastructure | Under the customer's own terms with Meta. Meta's role (processor for the customer, or independent controller) follows Meta's terms. |
| TikTok | Receiving and answering TikTok direct messages and comments; publishing | TikTok user IDs, message and comment content | TikTok's servers; its privacy policy names Singapore, Malaysia, Ireland and the United States | Under the customer's own terms with TikTok |
| Google (YouTube), LinkedIn, Telegram | Publishing the customer's posts to its own channels | Post content; no end-customer data | Provider's infrastructure | Under the customer's own terms with each platform |
3. AI providers (when the AI assistant or an AI feature is on)
| Provider | Purpose | Data | Location | Transfer mechanism |
|---|---|---|---|---|
| Anthropic | Writing the assistant's replies; reading images and documents sent by end customers | Message text, recent conversation, knowledge base excerpts, images and documents | Stored in the United States; Anthropic may route processing through the United States, Europe, Asia and Australia | Anthropic's data processing addendum, part of its Commercial Terms, with SCCs (Modules Two and Three) |
| OpenRouter | Routes requests to the model providers below; fallback replies | Same data as the model it routes to | United States | OpenRouter's data processing agreement, incorporated into its terms, with SCCs (Module Two). It passes data on to the model providers under their own terms; nashir restricts which providers it may use (see the training table below). |
| Google (Gemini models, reached through OpenRouter) | Transcribing voice notes, reading documents, expanding search queries, matching product photos | Voice notes, documents, images, the customer question being searched | Any country where Google or its agents maintain facilities | nashir has no direct agreement with Google for these models; the data reaches Google through OpenRouter (see the OpenRouter row) |
| OpenAI (direct and through OpenRouter) | Search embeddings for the knowledge base and for end-customer questions; describing images; comment moderation; fallback replies; live voice on the website widget | Customer questions, knowledge base text, images, comment text; live audio when web voice is used | OpenAI's infrastructure; its public API pages name no default location | OpenAI's data processing addendum, incorporated into its services agreement |
| Microsoft (Azure AI Speech) | Speech recognition and speech output for the WhatsApp voice-call add-on | Live call audio and the reply text | United Arab Emirates (Azure region UAE North) | Microsoft Products and Services Data Protection Addendum (version of 1 November 2026). It is incorporated by reference into Microsoft's Product Terms and applies without a separate signature; the Product Terms list Azure Speech as a core online service; transfers out of the EEA are under the 2021 SCCs. |
Use of API data for model training
From each provider's public terms, read on 2026-11-12.
| Provider and product | Used for training by default? | Retention stated | Source and page date |
|---|---|---|---|
| Anthropic — Claude API (Commercial Terms) | No. Anthropic may not train models on customer content from the services. Exceptions: feedback the customer explicitly submits, or an opt-in. | Inputs and outputs deleted within 30 days, longer where needed to enforce its usage policy or by law | https://www.anthropic.com/legal/commercial-terms (effective 2025-06-17); https://privacy.claude.com/en/articles/7996868 |
| OpenAI — API platform, including audio and realtime | No, unless the customer opts in | Abuse-monitoring logs up to 30 days | https://developers.openai.com/api/docs/guides/your-data (no date shown) |
| Google — Gemini API, paid services | No. Prompts and responses are not used to improve Google's products. | Logged for a limited period for abuse detection; no number given | https://aistudio.google.com/docs/terms (effective 2026-03-23) |
| Google — Gemini API, unpaid services | Yes, and human reviewers may read the content | Not stated | Same page |
| OpenRouter — its own processing | No, unless the account opts in to prompt logging. Logging is off on nashir's account (confirmed in the account settings, 2026-11-12). | Prompts and responses are not stored unless the account opts in; metadata is stored | https://openrouter.ai/privacy and https://openrouter.ai/terms (2026-08-31) |
| OpenRouter — requests routed to Google and OpenAI models | No. Routing is restricted to providers that OpenRouter lists as not training on inputs. On nashir's account all training and prompt-publishing endpoints, the 1% data discount, and input and output logging are off (confirmed in the account settings, 2026-11-12). The settings apply to every request made with the account, including requests from the automation workflows. The providers that serve the models in use are Google AI Studio, Google Vertex, OpenAI, Azure and Amazon Bedrock, each listed as not training (read 2026-11-12). | Set by the upstream provider. OpenRouter lists Google AI Studio as keeping prompts for 55 days and OpenAI as keeping prompts; it lists Google Vertex, Azure and Amazon Bedrock as not keeping them. | https://openrouter.ai/docs/features/provider-routing and OpenRouter's public provider list (read 2026-11-12) |
| Microsoft — Azure AI Speech (speech to text, text to speech) | Microsoft publishes no express statement on training for standard speech. Its Data Protection Addendum limits its use of customer data to providing the service and the business operations it lists. Real-time audio and text are not retained or stored. | None for real-time processing | https://learn.microsoft.com/en-us/azure/foundry/responsible-ai/speech-service/speech-to-text/data-privacy-security (2026-03-31); Microsoft Products and Services Data Protection Addendum (2026-11-01) |
On the official EU-US Data Privacy Framework list, read on 2026-11-12, Cloudflare, Google and Microsoft are active participants and Resend is listed as active with its re-certification under review. Anthropic, OpenAI, OpenRouter and Zoho are not active participants. Anthropic's data processing addendum (https://www.anthropic.com/legal/data-processing-addendum, effective 2025-02-24) is part of its Commercial Terms and includes standard contractual clauses, Modules Two and Three. OpenAI's data processing addendum (effective 2026-01-01) is incorporated into its services agreement.
With the AI assistant off, one path still reaches OpenAI and possibly Google Gemini: a text message sent through the nashir API can trigger a product lookup on the end customer's last inbound message. The customer switches this off for a send by setting skip_presentation to true in the request. See the Data Processing Agreement, Annex I, section 6.
4. Notifications and email
| Provider | Purpose | Data | Location | Transfer mechanism |
|---|---|---|---|---|
| Resend | Sending email to the customer's team (account email, alerts) | Team members' email addresses; alert emails can contain an end customer's name and phone number (for example a new booking) | United States | Resend's data processing addendum, part of its terms, with SCCs |
| Zoho | Hosts the nashir.ai mailboxes (support@) | Emails the customer sends to nashir, which can contain customer data | United States (Zoho's US data centre) | Zoho's data processing addendum, signed on 2026-11-12 |
| Telegram | (a) Alerts to the customer's own Telegram chat when it links one. (b) nashir's internal operations alerts. | (a) End customer's display name, masked phone number and a message preview of up to 140 characters. (b) Error alerts; an end customer's phone number appears only masked (country code and last 2 digits). | Telegram's data centres. Telegram states a location (the Netherlands) only for users who sign up from the United Kingdom or the EEA | None. Telegram offers no data processing agreement. |
| Google (Firebase Cloud Messaging), and the push services of Apple, Google and Mozilla browsers | Push notifications to the team's phones and browsers | Sender name and a message preview, unless previews are switched off | Provider's infrastructure | Firebase Data Processing and Security Terms, which provide for SCCs and the Data Privacy Framework |
5. Regional partners
In some markets, a regional partner confirms customers' subscription payments and provides support to customers in that market, including refund requests. It has access only to customers in its own market and none to customers in other markets.
6. Store integrations (when the customer connects a store)
| Provider | Purpose | Data | Location | Transfer mechanism |
|---|---|---|---|---|
| Salla | Receives order notifications from the customer's store to send over WhatsApp; product import | Shopper phone number and message text, received from Salla | Salla is established in Saudi Arabia | Under the customer's own terms with Salla |
7. Payment providers (not subprocessors)
Wayl (Iraq), Paymob (Saudi Arabia), EasyKash (Egypt), Switch (legacy) and Binance Pay handle the customer's own subscription payments. They receive the account holder's billing details (amount, email; EasyKash also name and mobile number). They do not receive data about the customer's end customers, and nashir uses them as a controller of its own billing. They are listed for transparency.
8. Not used
- No third-party error tracking service.
- No advertising or product analytics service.
- Cohere was removed from the service in 2026.
9. Changes to this list
nashir gives at least 30 days' notice before a new subprocessor starts processing customer data, by updating this page and by email to account owners. Customers can object as described in section 6 of the Data Processing Agreement. nashir keeps a log of each notice sent.
Change log
| Date | Change |
|---|---|
| 2026-11-12 | Version 1.0 |
| 2026-11-18 | Version 1.1. No subprocessor added or removed. The API product lookup can now be switched off per send (skip_presentation). |