Miirage AI Ethics
and Responsible Use Policy

1. Scope and purpose

This policy applies to AI-enabled products and services supplied by Miirage Limited, including holographic AI avatars, conversational assistants, synthetic voice or visual content, audience interaction features, analytics, customer demonstrations, managed deployments and related software services.

It is intended to support responsible innovation, legal compliance, customer trust and the safe use of AI in retail, corporate, public-space, event, brand, education and experience environments.

This policy should be read alongside our Privacy Policy, Cookie Policy, Warranty Policy, Returns and Refund Policy, applicable order forms, software licence terms, support agreements, statements of work and customer deployment terms.

2. Our responsible AI principles

Miirage applies the following principles when designing, configuring, deploying and supporting AI systems:

• Consent and authorisation: we do not create or deploy a realistic avatar, voice clone or likeness of a real person without appropriate written permission or lawful authority.
• Transparency: audiences should be able to understand when they are interacting with an AI system, synthetic avatar, digital character or AI-generated content.
• Safety and robustness: deployments should be designed with appropriate testing, content controls, security measures and human oversight.
• Fairness and respect: AI systems must not be designed to exploit, demean, discriminate against or unfairly manipulate individuals or groups.
• Accountability: each deployment should have clear responsibilities between Miirage, the customer, site operator, content owner and any third-party provider.
• Privacy by design: personal data should be minimised, protected and processed lawfully, fairly and transparently.
• Contestability and escalation: customers, individuals and affected parties should have a route to raise concerns, request review or escalate suspected misuse.

3. AI transparency and audience disclosure

Where Miirage systems are intended to interact directly with people, customers must ensure that appropriate notices, signage, on-screen messages or verbal disclosures are used so that audiences understand the nature of the interaction.

For public-facing AI avatars, Miirage expects deployments to make clear that the audience is interacting with an AI-generated or AI-assisted system unless this is obvious from the context.

Where AI-generated or AI-manipulated image, audio, video or avatar content could reasonably be mistaken for real human communication, the customer and Miirage should agree suitable disclosure, labelling or watermarking measures in advance.

4. Real people, likenesses, voice and identity rights

Miirage does not knowingly create or deploy an AI avatar, synthetic voice, realistic likeness or digital double of a real person without appropriate written authorisation from the individual, rights holder, employer, estate, agent or other authorised representative.

Written authorisation

• The permitted use, territory, duration, platform, audience, language and commercial context should be clearly documented.
• Any extension, reuse, repurposing or material change should require additional approval where required by the relevant agreement.
• Customers are responsible for ensuring they have all rights, permissions and consents for names, images, voices, scripts, brand assets, product claims and personal likenesses they provide to Miirage.

Voice and image integrity

• Miirage does not scrape public images, video or audio to replicate a specific person without authorisation.
• Miirage does not support non-consensual voice cloning, impersonation or misleading digital doubles.
• Approved avatars should operate only within the agreed scope, persona, tone, content boundaries and deployment context.

5. Fictional characters, celebrities and historical figures

Miirage does not knowingly create or deploy unauthorised replicas of copyrighted fictional characters, protected brand mascots, celebrities, athletes, performers, public figures or rights-managed estates.

Where a proposed deployment involves a brand character, performer, public figure, estate, archive, celebrity likeness, trademarked identity or protected work, customers must provide appropriate licences, approvals and usage rights before production or deployment begins.

6. Prohibited and restricted uses

Miirage products and services must not be used for unlawful, unsafe, deceptive, exploitative or harmful purposes.

Unless expressly approved in writing following appropriate legal, ethical and technical review, Miirage systems must not be used for:

• non-consensual impersonation, voice cloning, deepfakes or identity replication;
• political disinformation, covert propaganda or misleading public-interest communications;
• fraud, scams, phishing, social engineering or deceptive sales practices;
• illegal discrimination, harassment, hate, extremist content, sexual exploitation or abusive content;
• surveillance, law-enforcement identification, biometric identification or biometric categorisation;
• emotion recognition or sensitive inference in public spaces, workplaces or education settings;
• medical, legal, financial, employment, immigration, insurance, education or credit decisions without a separate approved scope and legal assessment;
• targeting children or vulnerable individuals with manipulative or exploitative persuasive techniques;
• collecting personal data in a way that is hidden, unfair, excessive or inconsistent with applicable privacy notices.

7. Public deployments, cameras, microphones and audience analytics

Miirage systems may include cameras, microphones, sensors, touchscreens, audience detection, dwell-time analytics, voice interaction or other interactive features depending on the product and customer configuration.

Where such features are used, the customer and site operator are responsible for ensuring that appropriate notices, signage, staff training, permissions and privacy documentation are in place for the deployment environment.

• Miirage does not use facial recognition, biometric identification, emotion recognition or biometric categorisation as standard.
• Audience analytics should be configured to use anonymous or aggregated data wherever reasonably possible.
• Voice, text, touch and avatar interaction data should be collected only where necessary for the agreed service, support, safety, analytics or improvement purpose.
• Recordings, transcripts and logs should not be retained longer than necessary for the agreed purpose.
• Additional safeguards may be required for public venues, healthcare environments, education settings, workplaces, children’s environments or regulated sectors.

8. Children and vulnerable audiences

Miirage does not design AI systems to exploit children, vulnerable people or people in distress. Where deployments may be accessed by children or vulnerable audiences, customers must ensure the content, interaction design, moderation and data collection are appropriate for that setting.

• Age-appropriate content controls should be applied where relevant.
• Personal data from children should not be collected without a lawful basis and appropriate notices, consent or safeguards.
• Deployments should avoid manipulative pressure, inappropriate profiling or targeted persuasion aimed at children or vulnerable people.
• Customers should carry out additional risk assessments for schools, family venues, healthcare settings and public attractions where minors may interact with AI avatars.

9. AI outputs, content controls and human oversight

Miirage AI avatars are intended to operate within controlled, brand-safe and customer-approved boundaries. Depending on the deployment, safeguards may include pre-approved Q&A libraries, moderated knowledge bases, restricted topics, fallback responses, escalation routes, logging, monitoring and human review.

• Customers should approve the intended persona, tone, content categories, scripts, knowledge base and fallback behaviour before launch.
• AI systems should not be relied on as the sole source of truth for critical, regulated or safety-sensitive decisions.
• Where live or semi-autonomous AI responses are enabled, customers should agree suitable prompt restrictions, content moderation and escalation processes.
• Miirage may pause or disable a deployment if a serious safety, reputational, legal, security or ethical issue is identified.

10. Data protection, privacy and security

Miirage designs and operates its AI-enabled services with data minimisation, access control, security and privacy safeguards in mind.

Where Miirage processes personal data, the applicable roles and responsibilities should be documented in the relevant contract, privacy notice, data processing agreement or statement of work. Depending on the context, Miirage may act as a controller, processor or joint controller.

• Personal data should be processed only for lawful, fair and transparent purposes.
• Access to customer data should be limited to authorised personnel and service providers with a legitimate need.
• Customer content, interaction logs, transcripts and diagnostic data should be handled in accordance with the applicable agreement and privacy policy.
• Miirage does not sell personal data collected through customer deployments.
• Miirage does not use identifiable customer deployment data to train third-party general-purpose AI models unless expressly agreed in writing.

11. Training data, customer content and intellectual property

Customers are responsible for ensuring that any materials they supply to Miirage are lawful, accurate and properly licensed. This includes scripts, FAQs, datasets, photographs, video, audio, brand assets, 3D models, product claims, trademarks, celebrity rights and personal likenesses.

Miirage may use customer-approved content to configure, test, operate and support the relevant customer deployment, but ownership and usage rights should be governed by the applicable order form, statement of work or licence terms.

Miirage may use anonymised, aggregated or non-identifying operational information to improve system reliability, diagnostics, security and service performance, provided this does not reveal confidential customer information or personal data in an unlawful way.

12. Third-party AI providers and integrations

Miirage products and services may incorporate or connect with third-party services, including cloud hosting, AI model providers, speech-to-text services, text-to-speech systems, translation tools, analytics services, CRM tools, content platforms, customer APIs and payment or booking systems.

Where third-party systems are used, deployment responsibilities, data flows, limitations, acceptable-use rules and service dependencies should be documented in the applicable order form, privacy documentation, data processing terms or statement of work.

Miirage does not guarantee the continuous availability, accuracy, performance or policy position of third-party AI models, APIs, software or platforms unless expressly agreed in writing.

13. Customer responsibilities

Customers play an important role in the safe and lawful deployment of Miirage AI systems. Unless otherwise agreed in writing, customers are responsible for:

• providing accurate, lawful and approved content for the deployment;
• securing all rights, consents, permissions, licences and approvals for supplied materials;
• ensuring the deployment context is lawful and appropriate for the audience;
• providing venue notices, signage, staff instructions and user-facing disclosures where required;
• ensuring compliance with sector-specific rules, advertising standards, employment rules, consumer protection rules and data protection requirements;
• not modifying, bypassing or disabling safety controls without Miirage approval;
• reporting misuse, suspected security incidents or harmful outputs promptly.

14. Risk assessment and higher-risk deployments

Some AI deployments may require additional legal, technical, operational or ethical review. This may include deployments involving public spaces, children, biometric data, healthcare, education, employment, financial services, political content, high-profile individuals, sensitive locations, government environments or regulated industries.

Miirage may require one or more of the following before proceeding:

• a written use-case assessment;
• a data protection impact assessment;
• a legitimate interests assessment;
• a customer risk assessment;
• additional security controls;
• enhanced human oversight;
• legal, brand, estate or rights-holder approval;
• specific contractual restrictions or an approved statement of work.

15. Monitoring, incidents and escalation

Miirage may monitor deployments for reliability, security, support, misuse prevention, content safety and contractual compliance where this is permitted by the applicable agreement and privacy notice.

Customers should promptly notify Miirage of suspected misuse, unauthorised access, unsafe outputs, privacy concerns, security incidents, unlawful content, impersonation concerns or audience complaints relating to an AI deployment.

Where appropriate, Miirage may investigate, recommend mitigation, apply content restrictions, update fallback behaviour, suspend features, disable a deployment or escalate the matter to the customer, rights holder, venue operator, regulator or legal authority.

16. Compliance with evolving AI regulation

AI law, standards and best practices continue to develop. Miirage aims to align its AI governance with applicable UK, EU and international requirements where relevant to the deployment, including principles of safety, transparency, fairness, accountability, data protection, human oversight and redress.

Where new legal requirements apply to a customer deployment, Miirage and the customer may need to update disclosures, technical controls, contractual terms, product configuration, record-keeping, documentation or operational processes.

17. Review of this policy

Miirage may update this policy as our products, services, risks, laws and industry standards evolve. The version published on our website will apply from the time it is made available, unless a separate signed agreement states otherwise.