What to Include in Your AI Policy

What Goes Into an AI Policy?

Start with what matters to you.

You don’t need a 50-page legal document. You need clarity. Here’s what actually matters:

Why You’re Using AI (Or Not)

Start honest. Are you using AI to speed up administrative work? Experimenting with it for content creation? Just trying to understand the hype? Write down all the ways you use AI. This grounds your entire policy.

More importantly, this forces you to define what you’re not using AI for. Where are human creativity and judgment non-negotiable? That boundary-setting is the heart of your policy.

Copyright and Whose Work This Is

This is the topic that keeps creative leaders up at night, and rightfully so.

Be explicit: if your organization creates something with AI’s help, who owns it? If you’re using AI trained on other people’s work, are you comfortable with that?

You probably shouldn’t be. Most AI models are trained on copyrighted material without sourcing, artist consent, or compensation. Others are trained on public sites like Reddit or Wikipedia, where there’s relatively little quality control or accuracy protection.

Here’s the practical part: label what’s AI-generated and what’s human-created. Be transparent with your audience. If you’re a cultural institution using AI to organize or analyze collections, say so. If you’re a creator using AI as a tool alongside your own work, say so.

And crucially: protect your community’s work. If you work with artists, tribes, or cultural practitioners, make it clear in your policy that their work won’t be fed into AI systems without explicit permission. This isn’t optional; it’s respect.

Traditional Knowledge and Cultural Sovereignty

If you work with indigenous communities, tribal knowledge systems, or culturally specific creative practices, this section is critical.

Your policy should explicitly protect against unauthorized use of cultural data, traditional knowledge, artistic styles, or sacred practices in AI systems. This means:

• Getting genuine consent before any cultural material enters an AI tool
• Being clear about what “consent” means in your context (it’s not a checkbox)
• Respecting tribal intellectual property rights and sovereignty
• Understanding that some knowledge isn’t meant to be automated or commercialized

This isn’t just ethical—it’s often legally required under tribal law and increasingly under national frameworks like the U.S. Native American Graves Protection and Repatriation Act (NAGPRA) and international guidelines.

You might even choose to publish your AI policy publicly and direct people to your Tribal Heritage Preservation Office or other cultural facility for approval of content usage.

Data Protection and Privacy

If you’re collecting any information about artists, audience members, community members, or staff, you need to be intentional about how that data interacts with AI.

Don’t feed personal data into third-party AI tools unless you absolutely have to. If you do, use tools with strong data protection agreements. Comply with GDPR if you work internationally. Be especially careful with sensitive cultural or personal information.

Human Review and Editorial Control

Every piece of AI-generated content your organization puts out should go through human eyes first.

This isn’t paranoia—it’s quality control and values control. AI hallucinates, perpetuates bias, and sometimes generates nonsense. You need someone (or a team) who can look at the output and ask: “Does this match our standards? Does it represent our community accurately? Would this hurt anyone?”

Transparency With Your Audience

If you’re using AI, say so. Label AI-generated content. Explain your reasoning.

This builds trust. It also gives your audience agency—they get to know what they’re engaging with and make their own judgments.

Watch for Bias and Fairness

AI systems inherit the biases of their training data and designers. A system trained mostly on Western art will “see” non-Western aesthetics as outliers. A model trained on a narrow demographic will fail for everyone else.

Your policy should commit to:

• Regularly testing AI tools for bias, especially around representation and cultural fairness
• Being willing to reject tools that don’t meet your standards
• Documenting what biases you’ve found and how you’re addressing them
• Bringing in voices from your community to evaluate whether something feels off
• Encouraging better prompting: ask AI to source outputs from respected institutions and provide links to those sources

Artistic Freedom Comes First

Be crystal clear: AI is a tool. It doesn’t replace human creativity, judgment, or voice.

Your policy should protect space for messy, experimental, human-centered creative work. It should also protect artists from having their style, voice, or work replicated without consent.

At the same time, smart engagement with AI matters. Catastrophizing with Terminator scenarios doesn’t serve anyone—it perpetuates fear at a moment when people should be learning how AI actually works.

Training, Questions, and Evolution

Your team probably doesn’t know much about AI yet. That’s normal. Build in time and resources for learning:

• Train staff on what your policy means relative to their daily work
• Create space for questions and concerns
• Plan to revisit and update the policy as AI changes (because it will)
• Make it clear that using AI responsibly is everyone’s job, not just IT’s

Environmental and Ethical Considerations

Training and running AI models uses enormous amounts of energy—and water. This disproportionately impacts Native and rural communities.

If sustainability matters to your organization (and it should), factor that into your choices about which tools you use and how often you use them.

Also: where is the money going? Who profits from the AI tools you choose? Are you comfortable with that?

And Remember…

Things are changing rapidly. You first draft will NOT be your final draft. Be comfortable with change.