Responsible use of AI

1. Purpose and limits

BadPigeon uses AI for two limited purposes: optional assistance with drafting or improving website wording, and automated checks intended to identify unlawful, harmful or prohibited text and images.

AI does not independently design or publish a website, decide whether a customer should subscribe, or make decisions that produce legal or similarly significant effects about a person.

2. Transparency and customer choice

Wording assistance is clearly identified as AI before it is used. Suggested wording is labelled as AI-suggested, and the customer may use it, edit it or disregard it. Customers remain responsible for reviewing the accuracy, suitability and rights implications of anything they publish.

Using AI wording assistance is optional. The core website-creation and management service remains available without accepting an AI suggestion.

3. Content checks and human review

An automated content check may prevent wording or an image from being published. Customers can usually continue by reviewing the highlighted material and making an appropriate change.

If you believe content was blocked incorrectly, use our verified contact form. Include the website address or draft, the affected section and why you believe the content is suitable. Do not include a password, verification code, authenticator code or recovery code.

A member of our team will examine the content and its context rather than simply repeat the automated result. We aim to acknowledge a request within two working days and complete most reviews within five working days. If a complex review takes longer, we will update you. We may allow the content, explain what needs to change or ask for further information.

4. Governance and safeguards

The Director of Beamond Limited is accountable for oversight of AI use within BadPigeon. The company maintains an inventory of AI uses and reviews their purposes, providers, information flows, risks, safeguards and human-review arrangements at least annually and before or after a material change.

• Only information relevant to the requested assistance or safety check is sent to the applicable provider.
• Instructions are designed to preserve supplied facts, avoid invented claims and treat customer content as untrusted data rather than instructions.
• Limited records of wording assistance, including provider, model, field, timestamps and acceptance, support monitoring and accountability without storing an additional copy of the suggestion.
• Provider failures do not silently approve unchecked content or replace an existing published website.
• Material incidents, recurring errors and provider or model changes are reviewed and may result in additional testing, safeguards or suspension of the feature.

More information about personal data, providers and retention is available in our Privacy notice.

5. Questions and concerns

For a content-review request, use the verified contact form. For privacy questions, use the same form and select Privacy or data-protection request.