AI governance
Which AI models power Tanqory features, which company runs each one and where, what they see, and the commitments that govern how we use them.
Updated
- Models Tanqory trains on your data
- None
- Model training on what we send to OpenAI and Anthropic
- Off
- ISO/IEC 42001, planned for FY2027
- Not certified
- SearchOpenAI embeddings, MongoDB Atlas
- AssistantOpenAI, Anthropic or DeepSeek
- WritingProduct and store drafts
- VoiceGoogle Gemini text to speech
- ImagesPhotoRoom, processed in the EU
Your data and training
Tanqory trains no AI models of its own. Every AI feature calls a model from a provider through its API, and what we send is used to answer that request.
With OpenAI and Anthropic, training on what we send is switched off on our account. DeepSeek is a model you can choose for the assistant; it runs in the People's Republic of China, where no transfer safeguard is in place for it yet, so choose it only after agreeing to that transfer. Apple's on-device models run on your phone and send nothing to Apple.
Where AI is used
- SearchProduct search and recommendations, from OpenAI embeddings in a MongoDB Atlas vector index.
- AssistantConversational answers over your help content and store data, on OpenAI, Anthropic or DeepSeek, through CopilotKit.
- WritingProduct descriptions and store content drafts.
- VoiceSpoken replies through Google Gemini text to speech.
- ImagesBackground removal through PhotoRoom, processed in the EU.
Our commitments
From our AI Ethics Framework and EU AI Act notice. These are published policies; they have not been audited.
- People decideA person reviews any automated decision that affects your rights, and you can appeal.
- Risk classesEvery AI use is classed by risk before it ships.
- LabelsChatbots and generated content are marked as AI.
- IncidentsAn AI incident is reported inside Tanqory within 24 hours.
Frameworks
Tanqory OÜ in Estonia is our representative under the EU AI Act. Our compliance notice sets out our roles as provider and deployer and how we classify risk.
- ISO/IEC 42001AI management system.Not certified. Planned for FY2027
AI providers and where they process data
The AI companies on our subprocessor list. A new one appears here when it is added.
| Provider | What it does for us | Where |
|---|---|---|
| OpenAI, L.L.C.AI | Language models, embeddings and image generation | United States |
| Anthropic, PBCAI | Claude language models, a provider merchants can choose | United States |
| Hangzhou DeepSeek AI Co., Ltd.AI | Language models, a provider merchants can choose | People's Republic of China |
| Google LLCAI | Gemini text to speech for spoken replies, and Gemini Nano on Android devices | United States |
| Apple Inc.AI | Apple's on-device language models on iPhone | On the device |
| PhotoRoom SASAI | Background removal and product image editing on the images you upload | France, European Union |
| CopilotKit, Inc.AI | The in-product AI assistant, on CopilotKit Cloud | United States |
Policies and documents
- AI Ethics FrameworkThe principles, controls and governance that Tanqory applies to keep its AI safe, transparent and fair.Updated
- Algorithmic Transparency PolicyTanqory's standards for notice, explanation and appeal when algorithms and AI systems affect people.Updated
- EU AI Act Compliance NoticeHow Tanqory handles AI systems it provides or uses for customers in the EU, EEA and UK under the EU AI Act.Updated
- AI Data Usage PolicyHow data is collected, used, stored and deleted when it is used to build, test or run AI features on Tanqory.Updated
- Artificial Intelligence Policy (AI Policy)Principles and roles that govern the AI systems Tanqory builds, buys or runs, including AI features on the platform.Updated
Questions
Can I turn off AI features for my store?
Write to privacy@tanqory.com and we will stop AI features using your data. A switch in the admin is not available yet.
Still have a question?
For anything this page does not answer, write to privacy@tanqory.com.
More in the trust center
Found a security issue?
Write to security@tanqory.com with the steps to reproduce it. We acknowledge every report within 24 business hours, keep you posted while we fix it, and never take legal action against research done in good faith under our disclosure policy.