AI Usage Policy – Our Approach to AI at O’Brien Media

Artificial intelligence can help businesses work more efficiently, improve accessibility and explore new ideas. However, AI-generated information can be inaccurate, incomplete or biased, and using AI can introduce privacy, copyright and security risks.

At O’Brien Media, we use AI carefully, transparently and with human oversight. AI supports the work of our team; it does not replace our professional judgement, creativity, responsibility or direct relationship with our customers.

How we may use AI

We may use approved AI tools to assist with tasks such as:

• Researching and organising information.
• Producing initial ideas, outlines or draft content.
• Reviewing spelling, grammar, clarity and accessibility.
• Assisting with coding, testing and technical troubleshooting.
• Summarising documents, notes or technical information.
• Creating or editing images and other creative material.
• Identifying patterns in website, marketing or operational data.
• Improving internal processes and customer support.
• Providing optional tools that help team members work in ways suited to their individual needs.

AI-generated work is treated as a starting point or supporting resource. A member of our team remains responsible for work delivered to a customer.

Human review and responsibility

We do not assume that AI-generated information is correct.

Where AI contributes to customer work, important outputs are reviewed by a suitably experienced team member. The level of checking depends on the potential consequences of an error.

This may include:

• Checking factual claims against reliable sources.
• Reviewing code before it is used on a live website.
• Checking content for accuracy, tone, accessibility and suitability.
• Reviewing images for misleading, inappropriate or unintended details.
• Testing changes before they are introduced into a production environment.
• Obtaining customer approval where this would normally form part of the service.

AI will not be given final authority over significant decisions affecting customers, team members or other individuals without appropriate human involvement.

Personal, confidential and customer information

We handle personal information used with AI in accordance with our Privacy Policy and applicable UK data protection legislation.

We follow these principles:

• We only use personal information where there is a clear and lawful reason.
• We limit information to what is reasonably necessary for the task.
• We do not knowingly enter passwords, payment card details, security credentials or similarly sensitive information into general-purpose AI tools.
• Confidential customer information is only used with an AI service where the use is appropriate, authorised and subject to suitable privacy and security controls.
• Where practical, information is removed, anonymised or pseudonymised before it is processed.
• We review how AI providers store information, where it is processed and whether it may be used to train their models.
• We prefer business services that provide appropriate contractual, privacy and data-retention controls.
• We carry out a data protection impact assessment where an intended use of AI is likely to create a high risk to people.

We do not sell customer information to AI providers.

We do not intentionally allow confidential customer information or personal information submitted through approved business AI services to be used for general model training unless this has been clearly assessed, lawfully authorised and appropriately communicated.

Automated decisions

O’Brien Media does not currently use AI to make solely automated decisions about customers or team members that produce legal or similarly significant effects.

If this changes, we will assess the use before it is introduced, provide appropriate information to affected people and apply the safeguards required by data protection law. These may include the ability to obtain human intervention, provide additional information and challenge a decision.

We will take particular care where a system could affect recruitment, employment, eligibility, access to services, pricing or another significant outcome.

Accuracy, fairness and bias

AI systems can reproduce inaccuracies and unfair assumptions found in their source material or training data.

We take reasonable steps to identify and reduce these risks by:

• Considering whether AI is suitable for the task.
• Checking important information against trustworthy sources.
• Reviewing outputs for potentially discriminatory or unfair content.
• Considering how an output may affect disabled, neurodivergent or otherwise underrepresented people.
• Avoiding AI systems that rely on unnecessary profiling or sensitive personal characteristics.
• Providing a route for concerns to be raised and reviewed by a person.

We will not use AI for social scoring or biometric categorisation of customers or team members.

Copyright and ownership

AI-generated material can raise questions about copyright, licensing and ownership.

When using AI to support customer work, we take reasonable steps to:

• Avoid deliberately requesting close copies of identifiable copyrighted work.
• Check important content for potential infringement or improper attribution.
• Use appropriately licensed source material and reference images.
• Respect trademarks, brand guidelines and the rights of creators.
• Keep human creative input and review at the centre of work delivered to customers.
• Explain any material limitations where the legal ownership or exclusivity of an AI-generated asset may be uncertain.

Customers should tell us if a project must not include AI-generated images, text or other material or if AI-generated images, text or other material is being provided to us for use in a project.

Security and AI-enabled systems

We treat AI systems as part of our wider cyber-security responsibilities.

Before introducing an AI tool or feature, we consider matters including:

• What information and systems it can access.
• Whether its access can be limited to what it genuinely needs.
• Authentication, permissions and account security.
• Supplier and software supply-chain risks.
• Retention, logging, monitoring and deletion controls.
• Prompt injection, manipulated source material and data poisoning.
• The possibility of an AI system producing unsafe commands or revealing confidential information.
• How the service can be disabled or safely withdrawn.

Information taken from websites, documents, emails or other external sources is treated as untrusted. Instructions found within that material are not automatically followed by an AI system.

Our approach is informed by the UK Government’s Code of Practice for the Cyber Security of AI and the National Cyber Security Centre’s Guidelines for Secure AI System Development.

AI agents and actions

AI agents can do more than generate content: they may be able to search systems, modify files, send messages or carry out other actions.

We apply additional safeguards before allowing an AI system to act on our behalf. Depending on the potential impact, these may include:

• Restricting the systems and information the agent can access.
• Giving it the minimum permissions necessary.
• Requiring human approval before consequential or irreversible actions.
• Testing it in a controlled environment.
• Maintaining records of significant actions.
• Monitoring for unexpected behaviour.
• Providing a clear way to pause or stop the system.

AI will not independently publish customer content, send customer communications, make purchases or carry out other significant external actions unless the activity has been specifically authorised and appropriate safeguards are in place.

Transparency

We aim to be open about material uses of AI.

Where someone communicates directly with an AI assistant or automated service provided by O’Brien Media, we will make this clear unless it would already be obvious from the context.

We may also identify content as AI-generated or AI-assisted where this is legally required, where failing to do so could materially mislead someone or where disclosure is otherwise appropriate. This is particularly important for realistic synthetic images, audio or video.

We do not generally label routine uses such as spelling assistance, summarisation or background research where AI has not materially altered the meaning or nature of the finished work.

Tools and suppliers

AI tools must be assessed and approved before they are used for O’Brien Media work.

Our assessment may consider:

• The provider’s privacy and security terms.
• Whether submitted information is used for model training.
• Data storage and deletion arrangements.
• The location of processing and international data transfers.
• Access controls and account management.
• Reliability, known limitations and available documentation.
• The provider’s approach to security incidents.
• Whether the tool remains appropriate for its intended purpose.

Approval is reviewed when a service, its terms or its intended use changes materially.

Team awareness and responsible use

Team members using AI receive guidance appropriate to their role and the systems they use. This includes understanding AI’s capabilities and limitations, protecting information, checking outputs and recognising potential security threats.

No team member is required to use generative AI as a condition of receiving reasonable workplace support. We particularly value the ways in which optional AI tools may help disabled and neurodivergent team members work in ways suited to them.

Uses we do not permit

We will not knowingly use AI to:

• Deceive, impersonate or manipulate people.
• Produce fake customer reviews, testimonials or endorsements.
• Create unlawful, discriminatory, abusive or deliberately misleading material.
• Make significant decisions about people without appropriate safeguards.
• Circumvent security, privacy or intellectual-property protections.
• Conceal a material use of AI where disclosure is required.
• Present invented sources, statistics or claims as verified facts.
• Replace professional legal, medical, financial or other regulated advice.

Concerns, mistakes and incidents

If an AI-related output is found to be inaccurate or inappropriate, we will assess and correct it where reasonably possible.

Suspected privacy or security incidents involving AI are handled through our normal incident-management process. Depending on the circumstances, this may include limiting access, disabling a tool, investigating what happened, notifying affected customers and making any legally required reports.

Customers and members of the public can ask us to review a concern involving our use of AI. A person, rather than an automated system, will consider the concern.

Customer preferences

We understand that customers may have their own rules or preferences about AI.

If you do not want generative AI used for a particular project or type of information, please tell us before work begins. We will discuss whether we can meet that requirement and whether it affects the scope, timescale or cost of the work.

Customer feedback and the use of AI

Clear feedback helps us understand what you want, make the right changes and keep projects moving efficiently. We prefer customers to provide feedback in their own words wherever possible.

We understand that customers may use AI to help organise notes, improve spelling or turn ideas into a clearer message. However, AI tools can sometimes:

• Change the intended tone or meaning of feedback.
• Add requests, assumptions or technical requirements that the customer did not intend.
• Present suggestions as firm instructions.
• Expand the agreed scope of a project without making this obvious.
• Produce contradictory, impractical or unnecessarily complicated requirements.

This can lead to misunderstandings, additional work, delays and potentially higher costs for both O’Brien Media and the customer.

If you use AI to prepare, rewrite or substantially expand feedback or instructions sent to us, please let us know. Before sending it, please also check that it accurately represents what you want and remove anything the AI has introduced that you have not personally requested.

Processing and responding point-by-point to AI-generated feedback is not included in the scope of your project with O’Brien Media. If submitted, we may ask that it be consolidated into clear, human-directed feedback before work continues, or additional fees may be required for our team to review, process, and respond to AI-generated feedback.

Any material change to the project will still be discussed with you before additional chargeable work is undertaken.

We welcome AI-assisted communication where it genuinely helps a customer express themselves, including as an accessibility aid. Our aim is not to discourage its use, but to make sure that the final instructions reflect the customer’s own intentions and that everyone has the same understanding of the work required.

Keeping this policy under review

AI technology, regulation and best practice are developing quickly. We review this policy regularly and when there is a significant change to:

• The AI tools we use.
• The way we use AI.
• Relevant legislation or regulatory guidance.
• Identified security, privacy or ethical risks.

If you have questions about this policy or our use of AI, please email [email protected].

This document is regularly reviewed and was last reviewed on 16 July 2026.