Data and AI.

1. Overview

This page explains how Atelia uses data in its artificial intelligence and machine learning systems, what third-party AI providers we use, and the steps we take to protect personal data when AI is involved. It complements our Privacy Policy and Terms of Use.

We treat AI not as a marketing feature but as an operating discipline. We believe shoppers and brand partners deserve to know what is happening behind any recommendation, advisory, or score.

2. What data we use in our AI systems

Atelia's AI systems work on the following categories of data.

• Product catalogue data. Product names, descriptions, prices, images, fibre composition, care labels, and related metadata sourced from public retailer websites, partner brands, and partner data feeds.
• Fashion knowledge. Domain knowledge curated by our team, including how fibres behave, how garments age, and what occasions different fabrics suit.
• User queries and interactions. What you ask Atelia, what you click on, what you save, and similar interactions, where they are needed to generate a response.
• Aggregated and anonymised usage data. Patterns of how groups of users interact with the Service, with personal identifiers removed.

We do not knowingly use special category personal data (such as health, biometric, religious, or political data) in our AI systems.

3. Third-party AI model providers

Atelia is a multi-layer reasoning system. To build it, we use a combination of in-house models, retrieval systems, and frontier large language models provided by third parties. Our current AI sub-processors include, without limitation, OpenAI, Anthropic, and Google.

When we send data to a third-party AI provider in order to generate a response, we do so under a written contract that includes data protection obligations consistent with UK GDPR. Where the provider is outside the United Kingdom, we apply appropriate international transfer safeguards as described in our Privacy Policy.

Where reasonably possible, we use enterprise or API tiers that contractually prevent the AI provider from training their own models on our prompts or outputs.

4. Data minimisation and anonymisation

We follow a data minimisation principle when working with AI.

• We only send to AI providers the data that is needed for the specific task being requested (for example, the user's query and relevant product context).
• We avoid sending direct identifiers (such as names or email addresses) to AI providers where this is not strictly necessary.
• Where personal data must be included in a prompt, we pseudonymise or aggregate it where this can be done without breaking the user experience.
• We do not send payment information, government identifiers, or other sensitive identifiers to AI providers.

5. Use of data for model training and improvement

Atelia maintains a clear separation between using AI to deliver our service and using data to train new models.

• Operational use. Every shopper interaction may involve generating a response using AI. This is operational and necessary to deliver the Service.
• Improvement and research. We may use aggregated and, where possible, anonymised data to evaluate, debug, and improve our own systems. Where this is done at scale or involves training in-house models, we rely on legitimate interests as our lawful basis under UK GDPR, balanced against your rights and freedoms.
• Third-party training. Where we have a choice, we do not allow third-party AI providers to use our prompts or outputs to train their own foundation models. Where this option exists in their commercial terms, we select it.

6. How to opt out

If you do not want your interactions to be used for our internal research or model improvement, you may opt out by writing to privacy@atelia.ai. We will action your request within a reasonable period.

Opting out does not prevent the operational use of AI to deliver responses to your queries, which is necessary to provide the Service.

7. Governance and review

We treat AI risk seriously and apply governance practices appropriate for our stage and scale.

• Reema Raisinghani serves as our Data Protection Lead and reviews AI-related processing activities.
• We maintain a record of processing activities consistent with Article 30 of the UK GDPR.
• We assess high-risk processing through Data Protection Impact Assessments (DPIA) where required.
• We review our prompts, system messages, and outputs for safety, bias, and explainability.
• We provide a route for users to challenge any AI-driven decision affecting them, by writing to privacy@atelia.ai.

We are compliant with the data protection and AI principles set out by the UK Information Commissioner's Office (ICO) and we follow the UK government's emerging AI assurance and AI safety guidance. Where formal certification schemes become available and material to our work, we will pursue them.

8. Known limitations of AI output

Our AI systems can be wrong. Specifically:

• Recommendations may not match every shopper's taste, budget, or context.
• Fabric explanations are derived from composition data and general fashion knowledge, and may not perfectly describe a specific garment's behaviour.
• Price-fairness reasoning is contextual and may be biased by available market data.
• AI outputs are informational and should not be treated as professional advice, certification, or guarantee.

Where Atelia advises "best avoided," it is sharing an opinion based on its reasoning over the available evidence, not making an authoritative judgement.

9. Contact

If you have any questions about how we use data in our AI systems, please write to our Data Protection Lead.

• Email. privacy@atelia.ai or reema@atelia.ai.
• Post. ATELIA LIMITED, 167-169 Great Portland Street, London, England, W1W 5PF.