Security

Our approach is to make the risky moments explicit. The agent can answer and recommend freely. Anything that changes an order needs the shopper to confirm and prove who they are first.

Before the agent changes an order, it verifies the shopper with a one-time code sent to the email address on file for that order. It also asks the shopper to confirm the exact action before it runs.

• One-time-code verification before any order change.
• Confirmation before actions: the shopper sees what will happen and approves it.
• Idempotent actions: a retry or a double click cannot apply the same change twice.

A second model checks automated replies for quality and safety before they are sent. When the check fails or the agent is unsure, it does not guess: the conversation goes to a person on the merchant's team, with the context attached.

Automated replies are labeled as AI. A shopper can ask for a person at any time.

Handychat is built for GDPR and EU AI Act transparency. It is a limited-risk chatbot under Article 50 of the AI Act, so its main duty is to make clear that people are talking to an AI. It is not used for any high-risk purpose listed in the Act: no biometric identification, no emotion recognition, no credit scoring and no hiring decisions.

What we do today:

• AI disclosure: every website chat shows "You're chatting with an AI assistant." On WhatsApp, Instagram, Messenger and email, the first AI reply in a conversation says it's from an AI assistant.
• A person on request: shoppers can ask for a person at any time, and the conversation is handed off to the merchant's inbox.
• Grounded replies: answers are based on the merchant's own store data and knowledge.
• Verified actions: order actions such as canceling an order, editing an address or applying a discount need a verification code sent to the order's email or the order's WhatsApp number, and an explicit YES before anything changes.
• Retention under merchant control: merchants set data retention per agent.
• Access logging: access to personal data is logged.
• Telling shoppers: merchants who deploy Handychat should tell shoppers they are using an AI assistant, and the chat widget does this for them.
• Public paperwork: our Data Processing Addendum and subprocessor list are public.

Read the Data Processing Addendum

Access tokens for connected services, including Shopify and messaging channels, are stored encrypted with AES-256-GCM. Traffic to Handychat uses HTTPS. Rate limits protect public and per-store endpoints from abuse.

Every request and background job is scoped to a single merchant's store. Data belonging to one merchant is not returned to another.

Handychat runs on managed cloud services from the providers listed on the Subprocessors page. Incoming webhooks from Shopify and from messaging channels are verified by their signatures before we process them; unsigned or invalid requests are rejected.

Handychat handles Shopify's mandatory privacy webhooks: customer data request, customer redact and shop redact.

If we confirm a security incident that affects merchant or shopper data, we will investigate, contain it and notify affected merchants without undue delay. The notification terms will be finalized in legal review.

If you believe you have found a vulnerability, write to privacy@handychat.ai with the steps to reproduce it. Please give us a reasonable time to fix it before you share details publicly, and do not access data that is not yours.

None yet. Handychat does not currently hold third-party security certifications or audit reports. If that changes, we will list them here.