Artificial Intelligence (AI) AI Policy Workbook – June 2026

Introduction

This is Version 3 of a policy created to support cultural organisations across the UK. Version 1 was written in July 2024 as AI use increased across the sector workforce. Version 2 followed in November 2024, adding governance questions and drawing on sector feedback. This Version 3 (June 2026) reflects the significant developments in AI capabilities, regulation, and sector practice that have occurred since then. 

Future development of a three-layer structure

Version 3 has developed into more of an AI Policy Workbook.  We will continue to develop it throughout the Summer (2026) and it will form part of a three-layer structure, which will be introduced in our first AI leadership course, Responsible AI for Marketing Leaders and CEOs , this October.   This will include: First, the AI policy, which holds core commitments, principles, red lines and accountability. Second, the operational guidelines, where day-to-day practice lives. Third, the connected policies, such as data protection, HR, EDI, accessibility, publishing, sustainability and digital tools, which continue to govern their own areas with AI-specific content added where needed.

This reflects where AI governance is going: away from one static document that tries to solve everything and towards a clearer structure where principles, practice and existing organisational responsibilities can work together.

 The AI landscape has shifted substantially since November 2024. Key changes include: Arts Council England publishing its Responsible AI Policy and accompanying Practical Toolkit (2025); the Museums Association revising its Code of Ethics (October 2025) to include explicit AI provisions; The Audiences Agency's Let's Get Real: AI report (March 2026) calling for a collective sector voice; the UK Government's Copyright and AI consultation report (March 2026); and the rapid emergence of agentic AI tools that act autonomously on behalf of users. Version 3 addresses all of these. 

Versions 1 and 2

Versions 1 and 2 of this policy were produced by the Arts Marketing Association and Target Internet in development with the AI Sector Support Group. The organisations involved are: Arts Marketing Association; AIM - Association of Independent Museums; Black Lives in Music; Clore Leadership; Family Arts Campaign; Future Arts Centres; Independent Theatre Council; Jocelyn Burnham (aiforculture); Kids in Museums; Museums Association; Music Mark; One Dance UK; OutdoorArtsUK; The Audience Agency; The Space; UK Theatre.  

Version 1 of this policy was Informed by Codelabs Sample AI integration and experimentation policy - https://codehopelabs.com/sample-llm-policy , Cambridge University Generative AI Guidelines, BBC Generative AI Guidelines, Civil Service Guidance on Generative AI, National Lottery Heritage Fund – Digital Heritage Leadership Briefing on AI, Claude AI.  

Version 2 of this policy was informed by feedback from sector professionals and further research into sample policies emerging across the cultural and commercial sectors.   

Version 3

Version 3 has been expanded by the Arts Marketing Association before further consultation with the AI Sector Support Group throughout the summer of 2026. It draws on additional feedback and sample policies.  Version 3 is additionally informed by: ACE Responsible AI in Practice report and Practical Toolkit (2025); Museums Association Code of Ethics 2025; The Audiences Agency Let's Get Real: AI (March 2026); UK Government Copyright and AI Report (March 2026); Nolan Principles of Public Life (as referenced in ACE's AI Policy); BRAID (Bridging Responsible AI Divides) research programme at Goldsmiths/Alan Turing Institute. 

We have kept this policy at a governance level but have added questions to lead towards the practical guidance that teams need. Please use this policy as a starting point for your organisation and amend it to fit your specific context. Please share your working policy with: cath@a-m-a.co.uk

---

Using this policy

This policy has been designed as a jumping off point for Boards/CEOs and their teams. You will want to make it relevant to your organisation considering your purpose, stakeholders and context. 

You may want to set up an AI working group to develop this policy and keep things moving as you integrate AI into your team’s activity.  

You may want to consider where AI is already being used and start by addressing the policy areas that most impact this existing activity.  

Our draft policy has 16 policy areas for you to consider. We’ve minimised crossover in these where possible but things like team training are relevant in a number of areas.  

During the process you are likely to want to consider what AI technologies you want the policy to cover. This could include: 

• Content generation tools 
• Collections management systems 
• Visitor experience technologies 
• Administrative and operational systems 
• Marketing and communication tools
• Agentic AI tools 

Since late 2024, AI tools capable of taking autonomous actions, browsing the web, sending emails, booking resources, executing multi-step tasks, have entered mainstream use. These 'agentic' tools pose different governance challenges from text generators and require specific consideration. See Policy Area 15 below. 

 The ACE Responsible AI Practical Toolkit (2025) also recommends organisations map who AI projects directly impact and who has relevant expertise, before committing to tools or approaches. 

---

Core policy areas to consider

1. Responsible experimentation

Governance principle: 

We encourage staff to conduct responsible experiments aligned with our mission and values. 

All AI experimentation should be mapped to our organisational purpose before proceeding. We recognise that responsible experimentation includes the right to conclude that a particular AI tool is not suitable for our organisation at this time. 

Key questions to consider: 

• How do we know the proposed AI use advances our mission? 
• What risks need to be assessed before going ahead? 
• Who needs to be consulted before AI tools can be used? 
• When can team members go ahead without consulting others? 
• Do we want to record all AI experiments? How? 
• What success metrics do we want to establish?  
• Is there a case for NOT adopting a particular AI tool? How do we document that decision? 

---

2. Data protection and privacy

Governance principle: 

When using AI systems that process personal data, we will comply with all relevant data protection legislation, including the UK GDPR. We will be transparent about our use of personal data in AI and obtain explicit consent where required. Sensitive, confidential, or personal information should not be input into third-party AI systems without appropriate data protection safeguards. 

We recognise the growing risk of unsanctioned AI tool use by staff ('shadow AI'). We will take proactive steps to inventory AI tool use across our organisation and ensure that data handling practices are consistently applied regardless of which tools are used. 

Key questions to consider: 

• What data will the AI system process? 
• Where will data be stored and processed? 
• What consent will we need? How will we get it? 
• When might we need a data protection assessment? 
• What data are we confident to put into different AI systems e.g. does it change if the system is paid for and has settings to keep all information within the temporary chat?
• How do we monitor and manage unsanctioned AI tool use by staff? 
• What is our approach if a staff member inadvertently shares personal data with an AI tool? 

---

3. Intellectual property and cultural ownership

Governance principle: 

We respect the intellectual property rights of artists, creators, and communities whose works or cultural heritage may be used in AI training data or outputs. We will obtain necessary permissions and give appropriate attribution where it is within the scope of our control. We will carefully consider the cultural and ethical implications of using AI in relation to objects or knowledge of cultural significance. 

We will actively monitor developments in UK copyright law and AI. We will not assume that existing uses of AI training data are legally settled, and will seek legal advice where necessary. 

For organisations holding digitised collections: we will establish a clear position on whether and how our collections content may be used for AI training, and communicate this position publicly. 

Key questions to consider: 

• How/when do we assess if our AI use impacts culturally sensitive material? 
• When do we need to consult relevant community stakeholders? 
• How do we ensure appropriate attribution? 
• What safeguards are needed? 
• Do we hold digitised collections that AI developers might wish to train on? What is our policy on this? 

---

4. Human oversight and editorial control

Governance principle: 

While we may use AI tools to generate ideas, content, or analysis, we will not publish or act on AI outputs without human review and editorial control. All AI-generated content will be fact-checked and edited by staff to ensure accuracy, alignment with our brand voice, and adherence to our institutional values before publication. 

We extend this principle to any AI tool that takes actions on our behalf (agentic AI). No AI tool will be configured to take consequential actions, including communications with stakeholders, financial commitments, or changes to our systems, without defined human checkpoints and approval steps. 

Key questions to consider: 

• Who is responsible for reviewing AI outputs? 
• What quality criteria should be applied? 
• How do we document the review process? 
• What escalation procedures are needed? 
• For agentic tools: what actions are permitted without human approval, and what requires sign-off? 
• What is our process if an AI tool takes an unintended or incorrect action? 

---

5. Transparency and Accountability

Governance principle: 

We will be transparent about our use of AI systems both internally and externally. When publishing AI-generated content or using AI in visitor-facing applications, we will clearly label it as such. We will maintain audit trails of our AI usage and establish clear lines of accountability, including nominating senior responsible owners for AI projects. 

We align our transparency commitments with the Nolan Principles of Public Life (selflessness, integrity, objectivity, accountability, openness, honesty, and leadership). Where AI has been used in funded projects, we will be transparent with funders about the nature and extent of that use. 

Key questions to consider: 

• What can we reasonably audit? 
• How far back do we go in the audit trail? 
• Where is the line on what we label as AI generated? 
• What if AI was used to generate ideas as one part of the process, do we label that? 
• How do we communicate our AI use to funders? 
• Do we publish a public statement on our AI approach?

---

6. Resource Management

Governance principle: 

AI deployment must be resourced appropriately and sustainably. 

We recognise that AI adoption must be proportionate to our size, mission, and capacity. We will not adopt AI tools simply to keep pace with sector trends. We commit to investing in the staff time and training needed to use AI responsibly, recognising that unresourced adoption creates more risk than value. 

Key questions to consider: 

• What budget is required? 
• What staff training is needed? 
• How do we measure ROI? 
• Is our AI adoption proportionate to our capacity? Are we at risk of over-adopting? 

---

7. Monitoring bias and fairness

Governance principle: 

We recognise that AI systems can perpetuate or amplify biases present in training data and design. We will take proactive steps to identify and mitigate biases using human oversight. Where we develop our own AI systems we will endeavour to use training data that reduces the risk of bias. We will be alert to the risk of AI-generated content creating a misleading or unbalanced interpretation of art, history, or culture. 

We are particularly alert to biases that could result in the misrepresentation of marginalised communities, the reinforcement of colonial interpretations of our collections, or the erasure of non-dominant cultural narratives. Our bias monitoring will include consideration of these specific risks alongside general algorithmic bias concerns. 

Key questions to consider: 

• What human checks can we put in place to consider bias? 
• How far can we impact this with the tools and platforms we choose? 
• Do we need to consider this in content coming in from external sources?
• How are we monitoring for cultural bias in AI-generated content about our collections or programmes? 

---

8. Social impact and job displacement

Governance principle: 

We will strive to use AI in ways that promote cultural understanding, inclusion, and accessibility. We will be mindful of the potential impact of automation on our workforce and commit to supporting staff in developing the skills needed to work effectively with AI. 

We extend our commitment to the wider creative workforce with which we work, including freelancers and creative partners. We will consider the potential impact of our AI adoption on the livelihoods of the artists and creatives we commission, and will engage openly with them about how we are using AI. 

Key questions to consider: 

• Do we need an AI working group? 
• How do we plan the development of skills across the team? 
• How does our AI use affect the freelancers and creatives we commission? 
• Have we communicated our approach to AI to our creative collaborators? 

---

9. Environmental Sustainability

Governance principle: 

Recognising the potentially significant environmental footprint of AI, we will aim to use AI efficiently and avoid unnecessary computational waste. We will give preference to AI providers with strong environmental credentials and sustainable practices. 

We will avoid using AI tools for tasks that can be done as effectively without them. Where AI tools are necessary, we will prefer smaller, more efficient models and providers with verified environmental commitments. We will include AI's environmental footprint in our broader sustainability reporting where data is available. 

Key questions to consider: 

• How do we assess which AI providers to work with? 
• How can we measure any additional environmental footprint? 
• How do we balance taking up the opportunities of AI with our environmental impact policy? 
• Are we using AI for tasks where it provides genuine efficiency, or simply because it is available? 

---

10. Stakeholder engagement and ethical review

Governance principle: 

We will proactively engage with our audiences, local communities, cultural stakeholders, academic experts, and policymakers to inform our approach to AI. Where appropriate, we will establish an ethical review process to assess AI projects and ensure they align with our values and legal obligations. 

We will contribute, where we are able, to sector-wide efforts to shape responsible AI policy in the cultural sector, including engagement with funding bodies and their evolving AI frameworks and relevant DCMS consultations. 

Key questions to consider: 

• Which parts of our AI use should be informed by stakeholders? 
• Who needs to input into designing the ethical review process to reduce bias? 
• Who needs to be involved implementing in the ethical review process? 
• Are we contributing to sector-level conversations about AI governance? 

---

11. Artistic freedom and human creativity

Governance principle: 

While recognising the creative potential of AI, this policy affirms the enduring importance of human creativity and artistic expression. We commit to using AI to complement and enhance human creativity, not replace it. 

We acknowledge that generative AI is being used by artists as a creative tool and that this has legitimate artistic value. Our commitment is to artist agency: ensuring that where AI is used in creative work we commission or produce, the human creative voice remains primary and artists retain control of how AI is applied in their practice. 

Key questions to consider: 

• How will we use AI in our creative/artistic work? 
• How will we do this with our partners/artists? 
• Where do we draw the line on what enhances human creativity and what replaces it? 
• How do we ensure artists working with us retain agency over their use or non-use of AI? 

---

12. Collaboration and knowledge sharing

Governance principle: 

Given the complex challenges posed by AI, we encourage collaboration and knowledge sharing with other arts and cultural organisations, academic institutions, tech providers, policymakers, and civil society. This includes participating in sector-wide initiatives to develop AI ethics guidelines, share best practices, and advocate for responsible AI policies. 

We encourage staff to draw on freely available sector resources including the ACE Responsible AI Practical Toolkit, CultureHive's AI resource library, and the Let's Get Real: AI report. We will share our own learning with the sector where possible, including via CultureHive. 

Key questions to consider: 

• Are there existing stakeholders/partners/communities of practice that we should be engaging with? 
• Who will lead on this work and how will they manage the additional workload? 
• Are we drawing on, and contributing to, sector resources to support knowledge sharing? 

---

13. Quality assurance of AI-generated content

Governance principle: 

We are committed to ensuring that any content created with generative AI is of the highest quality. This includes: 

• Rigorous Review Process: All AI-generated content will undergo a stringent review and quality assurance process to ensure it meets our standards for accuracy, relevance, and artistic integrity. 
• Alignment with Organisational Values: Content produced using AI must align with our organisation's values, mission, and strategic goals. 
• Continuous Improvement: We will regularly assess and refine our AI systems and processes to maintain and enhance the quality of AI-generated content. 
• Training and Guidelines: Staff will be trained on best practices for using AI tools to create high-quality content, and clear guidelines will be established to support this goal. 

Key questions to consider: 

• Who is responsible for developing and implementing e.g. quality assurance process? 
• How do we check our AI work aligns with our strategic priorities? 

---

14. Use of AI in Recruitment

Governance principle: 

We are committed to using AI in recruitment processes responsibly and fairly: 

• Bias Mitigation: We will implement measures to identify and mitigate biases in AI-driven recruitment tools to ensure fair and equitable hiring practices. 
• Human Oversight: AI tools will assist in the recruitment process, but final hiring decisions will be made by human recruiters to ensure a holistic evaluation of candidates. 
• Transparency: Candidates will be informed if AI tools are being used in the recruitment process, and we will provide explanations of how these tools influence decision-making. 

We will publish a clear position on applicant use of AI in applications. We recognise that prohibiting AI use entirely may disadvantage applicants with certain access needs, and will take a considered approach that focuses on authentic representation of skills and experience. 

Key questions to consider: 

• Are there other organisations already doing this? 
• Do we need to consider our stance on applicants using AI to write their applications? 
• How does our recruitment AI policy interact with our equity and inclusion commitments? 

---

15. Agentic AI and autonomous systems

Governance principle:

We will apply heightened scrutiny to any AI tool that acts autonomously on our behalf. Before deploying any agentic AI tool, we will define the scope of its permitted actions, establish clear human oversight checkpoints, and assess the risks of unintended or erroneous actions. No agentic AI tool will be permitted to take consequential external actions, including communications with audiences, funders, or stakeholders, or any financial transaction, without explicit human approval. 

We will treat AI agents as requiring defined permissions and audit trails, in the same way we would manage any system with access to our data and external communications. 

Key questions to consider: 

• Are any AI tools we currently use or are considering taking autonomous actions on our behalf? 

• What is the scope of actions any agentic tool is permitted to take? 

• What human checkpoints are in place before consequential actions are taken? 

• How do we audit what actions an AI agent has taken on our behalf? 

• What happens if an AI agent takes an unintended or harmful action? Who is responsible? 

• How do we communicate to audiences and stakeholders when they may be interacting with an AI agent? 

---

16. AI use in funding applications

Governance principle: 

We will be transparent with funders about how AI has been used in the preparation of funding applications, in line with funder requirements. We recognise that generative AI tools create potential risks around bias, transparency, data protection, and the rights of creators, and will use such tools in ways that align with existing legal obligations and our values. 

Where our organisation is involved in assessing applications or nominations (e.g. for prizes, commissions, or funding), we will establish a clear policy on whether and how AI tools may be used in assessment, and will be transparent about this with applicants. 

 Key questions to consider: 

• Does our use of AI in funding applications comply with the requirements of relevant funders? 

• How do we ensure that AI-assisted applications authentically represent our organisation's voice and vision? 

• If we assess applications or nominations, do we have a policy on AI use in assessment? 

• How do we communicate our approach to applicants? 

---

Additional Things to Consider 

Roles and Responsibilities

You may want to outline these. For example:

• Board: Overall governance oversight 
• Senior Management: Strategic implementation 
• Department Heads: Operational implementation 
• AI Champions: Day-to-day support 
• All Staff: Responsible use 
• Responsible AI Lead: Named individual (or shared role) with responsibility for keeping the policy current, monitoring sector developments, and supporting staff. 

Additional Documents

Depending on your choices in developing the policy, you may need to develop:

• AI usage register 
• AI Project Assessment Template 
• Training plan 
• Quality Assurance Checklist 
• Data Projection Checklist 
• Success Metrics Framework 
• Agentic AI Permissions Register: documenting which tools are authorised to take actions, the scope of those actions, and audit records 
• Public AI Statement: a brief public-facing statement of your organisation's approach to AI  

---

Key sector resources

The following resources are recommended starting points: 

• ACE Responsible AI in Practice report and Practical Toolkit (2025) Seven guides to building responsible AI policies, developed over a twelve-month research and development period. 

• ACE Public Position Statement on AI in Grant Making (2025) Sets out ACE's expectations for applicants and grantholders using AI in applications. 

• Museums Association Code of Ethics 2025 Includes specific guidance on responsible and transparent AI use in museums. 

• The Audiences Agency: Let's Get Real: AI (March 2026) Report of a 12-month collaborative action research project involving 32 leaders from 16 UK cultural organisations. 

• CultureHive AI Collection Curated collection of AI guidance, case studies, and tools for the cultural sector. 

• UK Government Report on Copyright and Artificial Intelligence (March 2026)  Sets out the government's current position on AI and copyright, including plans for a Creative Content Exchange pilot. 

• BRAID (Bridging Responsible AI Divides)  AHRC-funded programme supporting cultural institutions to explore ethical AI use, in partnership with Goldsmiths and the Alan Turing Institute. 
• ICO Guidance on AI and Data Protection The UK's authoritative regulatory guidance on how data protection law applies to AI systems. It covers best practice for data protection-compliant AI and how to interpret data protection law as it applies to AI systems that process personal data.
• Julie's Bicycle – AI and the creative industries (blog series) Julie's Bicycle has begun a blog series exploring AI use in the creative industries, covering environmental and ethical concerns, how the sector is responding, and what the future of AI in the arts might look like.
• UK Government AI Adoption Plan: Creative Industries (June 2026) he plan identifies the main barrier to AI in the creative industries as not lack of interest but lack of capacity, confidence and infrastructure, particularly among microbusinesses, small firms and freelancers. It also sets out measures flowing from the March 2026 copyright report, including a taskforce on AI labelling to propose best practice so consumers understand whether content has been made using AI, and a working group on independent and smaller creative organisations to explore whether government should support their ability to license content.