All articles

AI chatbot privacy and the EU AI Act: an ecommerce merchant's checklist

Quick answer: Before launching an ecommerce AI chatbot, understand how it processes personal data and make its use of AI clear to shoppers. The GDPR governs personal data; the AI Act adds separate transparency obligations. Ask your supplier about data flows, responsibilities, retention and customer information. A general “GDPR compliant” label does not explain your store's implementation.

This article provides general information for customer-service and shopping-assistant chatbots. It is not legal advice. Requirements depend on the use case and the data involved. Seek appropriate privacy or legal advice for your setup. The guidance below draws on the European Commission and European Data Protection Board.

Map the path of a customer message

A question about a product's material may seem impersonal, but shoppers can put names, email addresses or private circumstances into a free-text field. Recommending public products does not, by itself, mean a conversation contains no personal data.

Ask your supplier to walk through a fictional conversation. What leaves the browser, which services receive it, where is it stored and what enters a contact request? Include technical logs in the discussion. Use invented details rather than a real customer's transcript.

Create a simple table of data categories, purposes, recipients and owners. Mark unknowns as questions. This makes the discussion about your implementation rather than a collection of general security terms.

Define purposes and legal grounds

The GDPR includes principles of data minimisation, purpose limitation and storage limitation. The Commission's guide to data-processing principles explains these foundations.

Separate product advice, contact requests and any additional use. Does someone asking for dimensions need to provide an email address? What happens to their details if they request a reply? Ask separately whether conversations are used for quality review, service improvement or model training. The word “AI” does not answer those questions.

Processing personal data needs an appropriate legal basis. Consent is one possible basis, not an automatic answer for every chatbot activity. Review the Commission's legal-grounds guidance and assess each purpose. Opening a chat or requesting help should not be treated as blanket permission for unrelated later use.

Confirm roles and the processing agreement

A controller determines the purposes and means of processing personal data. A processor acts on the controller's behalf and instructions. Roles follow the actual activities, as the EDPB's controller and processor guide explains.

Ask for a written allocation of responsibilities. Who handles customer messages, and for which purposes does the supplier use them? Where a supplier acts as a processor, the processing needs a GDPR-compliant agreement. Clarify subprocessors and how changes are communicated.

Check that the contract and technical description cover the same service. If the contract mentions storage but a separate model service generates answers, ask how that service fits into the processing chain. Resolve the gap before approval.

Make the AI interaction clear

Article 50 of the AI Act addresses transparency, including systems that interact directly with people. It requires information about AI interaction unless this is obvious in context. Information must be clear and provided by the first interaction. The Commission's AI Act Service Desk provides the provision.

A practical starting point is a visible introduction such as “I am this store's AI assistant.” This is an illustrative interface message, not a legal sign-off. Check whether a new visitor notices it on mobile and in every supported language.

Distinguish the provider's responsibilities from the deployer's. Ask the supplier to explain that division and review the actual notice in your store. A product assistant should not automatically be equated with every high-risk AI system: purpose matters. The Commission's AI Act overview explains the risk-based framework.

Keep AI disclosure and privacy information distinct

A short AI notice does not explain personal-data processing. Give shoppers an understandable route to appropriate privacy information. Depending on the situation, this includes purposes, legal basis, recipients and retention information.

The EDPB's guide to individual rights emphasises clear information and workable request handling. Agree how a request for access or deletion relating to a conversation will reach the responsible person. The rights that apply depend on the circumstances.

Rehearse the process without customer data. Where would a visitor find contact details? Who receives the request? Who can identify the relevant systems? These are useful questions to settle before a real request arrives.

Ask about retention, location and access

Request separate answers for conversations, contact requests, logs and backups. Ask how the agreed retention period is implemented and how that can be demonstrated. A contractual period and a working deletion process are separate things to verify.

Identify where processing occurs and who can access the data. An EU or EEA server location alone does not settle questions about model services or subprocessors. Transfers outside the EEA need separate assessment; see the EDPB's international transfers guide.

Ask for practical security information: how are permissions limited, customer data separated and incidents handled? The EDPB's security guide describes technical and organisational measures appropriate to risk. One general security statement is not an answer to every implementation question.

What can be said about Neuwon?

JR Soft supplies Neuwon. Its current security and privacy page describes controlled onboarding and the website's contact-data practices. It explicitly says it is not a complete legal privacy notice. Do not treat it as a full description of every merchant's chat processing.

Neuwon publication follows private preview and approval. Use that stage to review privacy information and AI disclosure as well. This article does not promise a particular storage country, deletion of all information after a fixed period or a model-training policy. Obtain those answers for your implementation and agreements.

Magento / Adobe Commerce and WooCommerce are available now. Neuwon also works on ordinary websites, and Shopify is in development. Platform choice does not remove the need to understand data flows. The WooCommerce onboarding guide complements this checklist with source preparation.

Frequently asked questions

Must shoppers be told they are talking to AI?

The AI Act's starting point is disclosure of AI interaction unless it is obvious in context. Prepare a clear notice at the first interaction and confirm the responsibilities for your implementation with the supplier.

Does an AI notice replace privacy information?

No. It explains the nature of the interaction. Privacy information answers different questions, including why personal data is processed and how individuals can exercise their rights.

Does every chatbot conversation require consent?

The legal basis depends on the purpose and circumstances. Consent is one possibility. Do not combine unrelated purposes or infer general permission simply because someone starts a conversation.

Is a supplier's GDPR claim enough?

Ask for an implementation description and the necessary agreements. Review data flows, roles, subprocessors, retention, access and request handling. A general marketing claim does not describe your store's complete setup.

Where should Neuwon privacy questions start?

Read the current security and privacy page and explain your planned use to JR Soft. Obtain implementation-specific answers before approval, especially where contact details or other personal data enter the chat.

Review the checklist before approval

Explore Neuwon's AI shopping assistant, check pricing and contact JR Soft. Bring your data-flow questions, proposed customer notices and open contractual points. An agreed operating model provides a clearer basis for approval than a general promise of easy automation. For preparation beyond ecommerce, read whether a company website needs a chatbot.