AI Statement

This statement explains how BrandedMark uses artificial intelligence (AI), what happens to your data when it does, and how we meet our obligations under the EU AI Act and data protection law. It forms part of your agreement with us and is read together with our data processing agreement and the Acceptable Use Policy.

In short: AI helps your team and your customers. Our AI provider never trains on your data. We learn from support questions, answers and satisfaction ratings to improve answers, with personal details and your company and brand names removed first, and you can switch this off. Your data is hosted in the EU (our AI provider may process prompts elsewhere, see section 3), we are adding notices that tell consumers when they're talking to AI, and you choose which AI features you use.

#1. Our principles

1. Your data stays yours. Our AI provider never trains on your data. Before we use support conversations to improve answers, we remove personal data and your company and brand names. What we learn from your product information improves only your own assistant, and we don't train or fine-tune any AI model shared between customers on your data. You can switch this off.
2. Humans stay in control. AI drafts, suggests and answers routine questions. Your team sets the rules and can step in at any time.
3. Transparency. We are adding notices so that people are told when they're interacting with AI (section 6).
4. Where AI processing happens. AI requests are processed by Google, which may process them outside the UK and EU under its data processing terms and standard contractual clauses. Everything we store stays in our EU hosting, apart from document-search indexes of product documents you mark Public and Include in AI.
5. Your choice. You decide which AI features you use, and we'll switch AI off entirely if you ask.

#2. AI features

Preview AI features. New AI features may be released as previews. They are labelled "Preview" in the product, off until an administrator switches them on, provided as is and outside the SLA, and not for production-critical use. They may change or be withdrawn on notice. We show any charge before you switch one on.

#3. Provider and location

• Model provider. Google's Gemini models. Google may process AI requests outside the UK and EU, and does not use your data to improve its products. Google is on our published sub-processor list.
• No training. Google is contractually prohibited from using your prompts, data or outputs to train or improve its models.
• Retention. Google keeps prompts and responses only for a limited period, to detect misuse. Document-search indexes of product documents you mark Public and Include in AI are held by Google (no committed location) and deleted when you delete the documents or leave BrandedMark.
• Service Improvement. Unless it's switched off for your account, we may learn from support conversations (the question, the answer and the satisfaction rating) to make answers more accurate. We remove personal data, and your company and brand names, first. What we learn from your product information improves only your own assistant. Across customers we change only the shared instructions, guardrails and checks, which never contain your content, and we don't train or fine-tune any AI model shared between customers on your data. Nothing is shared with our AI provider for training. It's on for accounts created after our Customer Terms take effect, and off for earlier accounts unless they switch it on. See Customer Terms clause 6.4.
• Our records. We keep a record of AI prompts and responses with your conversations, so you can review what the AI did. These traces follow the same retention rules as the conversations (2 years from last activity, or earlier deletion).
• Model choice. We choose and govern the models centrally. We may change the model or version to improve quality or safety. If we change provider, we'll give the sub-processor notice set out in the DPA.

#4. Human oversight and accuracy

AI can make mistakes. To manage that:

• the assistant answers from your own product data and documents, and only uses the tools and actions you enable;
• where you have set up handover to your team, it hands over when it's unsure, when a consumer asks, or when your rules require it;
• your agents can review, edit or take over any conversation, and AI-generated designer content stays a draft until you publish it;
• conversations can be quality-scored and reviewed; and
• we recommend keeping a human decision-maker for anything with financial or legal consequences, such as refunds, warranty rejections or safety information.

AI output is provided to support your team. You remain responsible for the content you publish and the decisions you make, including what the assistant tells your consumers in your name.

#5. Your controls

• Choose which AI features you use. AI support and AI image features are part of your plan. Email auto-reply and product chat can be switched off in your settings. To switch AI off for your whole account, contact [email protected] and we'll do it.
• Choose which tools and actions the assistant can use, including any external MCP servers you connect. Data sent to services you connect is governed by your agreement with them.
• Set when the assistant must hand over to a human.
• Delete conversations and documents. Deletion flows through to the AI indexes.
• Switch Service Improvement off, or on, by emailing [email protected] or in your Order Form.
• Choose whether to switch on Preview AI features.

#6. Transparency to consumers (EU AI Act, Article 50)

We are adding the following disclosures to the Service:

• a notice in the chat assistant that tells consumers they're talking to an AI, from the first message;
• a label on AI-written emails that are sent without human review, saying they are AI-generated; and
• a machine-readable AI marker on AI-generated text in the emails, messages and content data we produce.

Images generated or edited with our AI carry our provider's invisible watermark identifying them as AI-generated. The default consumer privacy notice we provide explains the AI processing.

Once these disclosures are in place, do not remove or obscure them. If you publish AI-generated text on matters of public interest, you are responsible for any disclosure the AI Act requires of you as deployer.

#7. EU AI Act

For the AI features in BrandedMark we are the provider under the EU AI Act, and when you use them with your consumers you are the deployer. Our AI features are not high-risk. Section 6 describes the transparency measures we are adding, and the people who build and run them follow our AI use policy and AI-literacy training.

#8. Your responsibilities as deployer

• Use AI features in line with the Acceptable Use Policy, and keep the Article 50 disclosures in place.
• Review AI-generated content before you rely on it for safety, legal, regulatory or contractual statements, including Digital Product Passport content.
• Don't configure the assistant to make solely automated decisions with legal or similarly significant effects on consumers without the safeguards the law requires.
• Make sure your own privacy notice covers your use of AI features. You can use our default consumer notice.
• Give your staff the training they need to use AI features responsibly (AI literacy).

#9. Prohibited uses

Don't use BrandedMark's AI features for anything prohibited by Article 5 of the EU AI Act, to make people think they're dealing with a human, or for anything our Acceptable Use Policy prohibits. That includes manipulating our AI features and systematically capturing how our AI features behave to build a competing model. You may use your own AI Output for anything else.

#10. As AI becomes more capable

AI is improving quickly. Future systems may match or exceed people at many tasks, which is sometimes called artificial general intelligence or superintelligence. However capable the models we use become:

• your team stays in control. AI acts only within the tools and permissions you give it, and you can switch it off;
• AI won't make decisions with legal or similarly significant effects on people without human review;
• we test a new model on our own features for safety and quality before we use it, and we tell you about material changes to how our AI features work; and
• we update this statement as the technology and the law develop.

#11. Contact

Questions about AI at BrandedMark: [email protected]. Security concerns: [email protected].

#Version history