ShopRise reads your catalogue, rewrites it, and publishes back to your storefront. That only works if you can see exactly what we touch, what we store, and what we will never ask for.
Nothing on this page is aspirational language. If a control is in progress, it says so.
Your catalogue, your content and your rankings live in an EU-hosted database, isolated per store. We store what we need to score and improve your storefront, and nothing else. No customer PII, no orders, no payment data ever enters our systems.
Least privilege by default. Every engineer authenticates with SSO and MFA, production access is scoped and time-bound, and each write back to your store is logged with the person who approved it in the Change Queue.
TLS 1.3 in transit and AES-256 at rest across the application database, backups and object storage. Shopify access tokens are encrypted with a separate key and are never written to logs or analytics.
We call model APIs under agreements that disable training on our inputs and outputs. We do not fine-tune models on customer content, and one merchant's catalogue never becomes another merchant's answer.
This is the fear every merchant has when they hand a catalogue to an AI tool: that the descriptions they spent years writing quietly end up inside a model that a competitor will use next quarter. So let us be exact about what happens.
ShopRise is a client of model providers, not a model builder. Your content travels to an inference endpoint, an answer comes back, and that is the end of the transaction. There is no fine-tuning step, no shared embedding pool, no "improve the product" checkbox that we quietly leave on.
Current status: zero-retention and no-training terms are in place with our primary model provider. Any provider we add is held to the same bar before it processes a single customer record.
Your product copy, brand context and rankings are sent as inference inputs. No customer content is used to train, fine-tune or evaluate a model, ours or anyone else's.
Our agreements with model providers exclude our inputs and outputs from model training. This is a contract term, not a setting we hope stays switched off.
Prompt and response logs are kept only long enough to debug a failed job, then deleted. Generated drafts live in your Change Queue, in your account, under your control.
Every store is a separate tenant with its own row-level isolation. One merchant's catalogue is never used as context, as an example, or as inspiration for another merchant's output.
One OAuth install, seven scopes, and a hard boundary around everything that has a person's name on it. This is the exact screen Shopify shows you before you click install.
read_products Read titles, descriptions, variants and images so we can score them and research the right keywords.
write_products Publish the title, description and metafield changes you approved in the Change Queue. Never before you approve.
read_content Read your existing blog articles to audit them and detect gaps against your competitors.
write_content Publish approved articles and article metadata back to your blog.
read_themes Read your theme templates to find missing structured data and render-blocking scripts.
write_themes Install the theme app extension that injects schema. No other theme file is touched.
read_online_store_pages Read static pages such as About and FAQ so the AI understands your brand context.
read_customers Customer names, emails and addresses. We never see them.
read_orders Orders, carts and fulfilment. SEO does not need your revenue lines.
read_payment_* Payment methods, cards and payouts. Out of scope by design.
Customer PII, orders and payment data are outside our permission set entirely. Not redacted, not filtered: never requested, so never received.
The complete list. If a vendor is not on it, it does not process your data.
| Subprocessor | Purpose | Region |
|---|---|---|
| Anthropic | Claude API for content generation and semantic analysis | US / EU |
| Vercel | Application hosting and CDN | EU + global edge |
| Supabase | Application database and authentication | EU |
| SEMrush | Keyword volume and SERP data | EU / US |
| Shopify | Store integration via the Admin API | Region of the merchant |
| Google Analytics 4 | Traffic and conversion attribution | EU / US |
| Resend | Transactional email | EU |
This list is kept current and reviewed every quarter. Customers on an active plan are notified by email before a new subprocessor begins processing their data, and can object. Last reviewed: July 2026.
ShopRise is a young product. Plenty of vendors of our age claim badges they have not earned. Here is the unedited version.
We process personal data as a processor on your behalf, under a data processing agreement. Transfers outside the EU run on Standard Contractual Clauses. A DPA is available on request and is included by default on Enterprise.
Our controls are documented and being implemented against the SOC 2 Trust Services Criteria. We are not SOC 2 certified today, and we will not say we are until an auditor has signed the report. We will publish it here when it exists.
Our information security policies are written against the ISO 27001 Annex A control set. There is no certificate behind that alignment yet, so treat it as posture rather than proof.
Access, rectification, erasure, restriction, portability and objection. Send a request to privacy@shoprise.ai and we respond within 30 days, or sooner if the request is straightforward.
We would rather hear it from you than from a customer. Good-faith research is welcome and will never be met with a lawyer.
security@shoprise.aiIf you make a good-faith effort to comply with this policy during your research, we will consider it authorised, we will not pursue or support legal action against you, and we will help make that clear to anyone who asks.
Good faith means: do not access, modify or destroy data that is not yours, do not degrade the service for other users, stop as soon as you have proof of a vulnerability, and give us a reasonable window to fix it before publishing.
Out of scope: volumetric denial of service, spam or social engineering of our staff, scanner output with no demonstrated impact, and missing best-practice headers that are not exploitable on their own.
Short answers, no hedging. If yours is not here, write to us and we will answer it in the same tone.
Run the free scan first. It uses public data only, so you can judge the output before you install anything.
No install required · Report within 48 hours
