AI Transparency & Ethics Policy

Our commitment

Meet George uses AI to help UK businesses compare and switch business energy supply more efficiently. We are designed to use AI in a way that supports safety, security and fairness, with human oversight and clear accountability. This policy explains how our AI-assisted features work, the safeguards we are designed to implement, and how you can raise concerns.

1. Scope and who this applies to

This policy applies to our UK business energy platform for businesses (including sole traders, limited companies, LLPs and trusts). Our service is intended for business users aged 18 or over. We do not knowingly collect personal data from children.

2. How we use AI

Our platform is designed to use AI to assist with:

• Document data extraction - reading bills and contracts to extract key details (such as meter identifiers, usage, rates, supplier details and contract dates).
• Assistance and Q&A - answering questions about uploaded documents and helping you locate and understand relevant clauses.
• Recommendations - helping you compare options by presenting relevant tariffs and explaining key factors.

AI is an assisting tool. You remain responsible for reviewing key information and confirming decisions.

3. Human oversight and no solely automated “significant” decisions

We are designed so that you must review and confirm key steps (such as submitting a switch or signing documents). We do not intend to make solely automated decisions that produce legal or similarly significant effects. Where we introduce automation that could materially affect eligibility or outcomes, we will signpost it clearly and implement appropriate safeguards (which may include additional human review, explanations, and user controls).

4. Explainability - “why am I seeing this?”

We are designed to provide clear, understandable reasons for outputs where it matters - for example the factors that contributed to a recommendation (such as estimated cost, contract structure, supplier terms, and your provided usage data). We aim to present explanations in plain language and avoid “black box” outcomes. Where an output is uncertain or depends on assumptions, we are designed to indicate that.

5. Fairness and suitability

We are designed to reduce the risk of unfair outcomes by:

• Testing and monitoring AI-assisted features for quality and consistency across common business scenarios.
• Avoiding the use of protected characteristics or sensitive personal attributes to alter pricing or personalise quotes.
• Designing recommendations to be driven by supplier terms and your inputs, not hidden model behaviour.
• Reviewing how options are presented (for example, ordering, labels and explanations) to reduce misleading impressions.

We do not use AI to set supplier prices. The price you see is the price offered by the supplier (subject to their terms, eligibility checks and availability).

6. Security and abuse resistance

Because AI systems can be targeted, we implement controls intended to protect the platform and users, including:

• Rate limiting and abuse monitoring to reduce automated attacks and scraping.
• Controls designed to detect and limit prompt injection, prompt hacking, model extraction attempts, and attempts to manipulate AI behaviour in a way that compromises security, privacy, or service integrity.
• Access controls, least-privilege permissions, and audit logging for administrative actions.
• Monitoring and alerting designed to identify unusual access or activity patterns.

If we identify a material AI-related security issue or integrity risk, we are designed to pause, restrict, or roll back the relevant feature while we investigate and remediate.

7. Data usage, privacy, and retention

Our Privacy Policy explains what personal data we process, why, and how long we keep it. In addition:

• Your documents and the extracted data are used to provide the service (quoting, switching support, document understanding and related operational steps).
• We are designed to minimise data sent to AI providers and to avoid sending unnecessary personal data.
• We are designed to delete uploaded documents in line with the retention periods in our Privacy Policy.
• We do not use customer content to build advertising profiles or to target advertising.

8. AI interaction analytics and publishing insights

To improve our AI-assisted features and service quality, we may collect analytics about how users interact with AI features (for example topic categories, feature usage, error rates, and aggregated patterns).

• We are designed to prefer aggregated and de-identified analytics over storing raw question text.
• Where we use examples in educational materials or blog content, they are illustrative and may be synthetic. We do not publish customer-identifiable queries or customer-specific confidential information.
• Where we retain any AI interaction logs for troubleshooting or security purposes, access is restricted and retention is limited.

9. Third-party AI providers and training restrictions

We use third-party AI providers via our infrastructure.

We are designed to prevent your personal data being used to train third-party AI models. Where we use third-party AI providers, we configure and contract with them to restrict training on customer content, and we select providers and settings designed for business use.

Where a provider offers “zero retention” or similar business controls, we are designed to use those settings where appropriate for the data being processed. (Some operational metadata may still be processed for security and reliability.)

Our Trust Centre lists our current vendors/sub-processors, including regions and the applicable international transfer mechanism: meetgeorge.co.uk/trust-centre

10. Testing, monitoring and improvement

We are designed to operate AI-assisted features with ongoing quality control, including:

• Routine evaluation of outputs for accuracy and reliability.
• Monitoring for recurring failure modes (for example extraction errors caused by poor document quality).
• Controlled rollout of material AI changes (for example staged releases) where practical.
• Tracking model and prompt versions for auditability and incident investigation.

We review this policy and our AI controls at least annually and when we make material changes.

11. Your responsibilities

To help the system work well, you should:

• Upload clear, complete documents where possible.
• Review extracted details for accuracy before you act on them.
• Ensure you have authority to upload documents and submit requests on behalf of the business.
• Avoid entering unnecessary personal data into free-text fields (for example bank details, personal addresses, or special category data) unless the service explicitly requests it for a defined step.

12. Accountability and how to raise concerns

Meet George is accountable for the design and operation of the platform. If you have concerns about an AI-assisted output or recommendation, contact us and we will investigate.

Contact: hello@meetgeorge.co.uk Privacy: privacy@meetgeorge.co.uk

Service complaints are handled under our Complaints Procedure (published on our website). If we cannot resolve a complaint within 8 weeks, or we issue a deadlock letter, eligible business customers may be able to escalate to the Energy Ombudsman under the Ombudsman’s scheme rules.

13. Updates to this policy

We may update this policy from time to time to reflect changes in our product, AI capabilities, or regulatory expectations. Where changes are material, we will post a notice on our website or in-product and, where appropriate, notify account holders.