Define the purpose and human responsibility
AI work needs an agreed purpose, approved information sources, acceptance checks and a named human owner. Outputs can be inaccurate, biased or incomplete. Review them before relying on them, and define when the system must stop or hand a task to a person.
AI output is not a substitute for regulated professional advice. Uses affecting health, employment, credit, legal rights or essential services need appropriate specialist assessment and meaningful human responsibility. We do not promise that AI is always accurate or suitable for every task.
Voice, likeness and disclosure
Use a person's voice, image or likeness only with the required rights and permission for the purpose, channel, territory and period. Consent to record is not automatically consent to clone a voice. A speaker's permission is also separate from the permission needed to contact a recipient.
We do not support deceptive impersonation, fake endorsements or testimonials, or non-consensual intimate content. Disclose synthetic content where law or platform rules require it and where needed to avoid misleading people. Do not remove required provenance information or labels.
Calls and conversations
Before outbound commercial calling, check recipient consent and preferences, telecom requirements, permitted calling resources, provider rules and opt-out handling. An AI consent checkbox does not replace those duties.
Identify the business and do not pretend the system is a particular human. Provide a suitable route to a person. Recording, transcription and retention need their own assessment, notices and controls.
We do not use voice systems for emergency response, unlawful spam, harassment, fraud or unauthorised political calls. A service enquiry is not general permission for promotional calls.
Data, tools and testing
Approve providers, connected systems, data locations, retention and training settings before use. Give the system only the access it needs. Material external actions, such as publishing content or changing records, need the approval process agreed for the project.
Test inaccurate outputs, attempts to manipulate instructions, privacy leaks, access limits, accessibility and handover failures. Keep an owner who can pause the system and investigate. A demonstration is not proof that every production risk has been resolved.
Claims, release and concerns
Support advertising claims with evidence and disclose paid or affiliate relationships where required. AI does not excuse false claims or missing rights. Agree release criteria, monitoring responsibilities and a response to failures.
Email legal@lucidgrowthmedia.com about a rights or content concern. Urgent incidents should use the project's agreed escalation route. This policy is a work standard, not a certification or guarantee of legal compliance.
Evidence, usage limits and suspension
Clients must supply evidence of rights to voices, likenesses, source content and recipient lists, and identify any restrictions on territory, duration, channel or reuse. We may require written releases and refuse work where permission cannot be established. A warranty from the client does not authorise deceptive content.
Agreed usage limits, provider charges, supported languages, human escalation, error handling and monitoring must be recorded before launch. Changes to models, prompts, data sources or tools may require renewed testing. A prototype does not create a promise of unattended operation.
We may suspend an affected system or release where misuse, missing consent, infringement or a security risk is reasonably suspected. Client approvals, third-party claim responsibilities, payment and liability follow the accepted agreement. No output is guaranteed to be unique, legally protectable or free of error.