Without a policy, each person makes their own decisions about AI. One team member pastes customer emails into a public chatbot; another publishes AI images without checking rights; a third avoids AI entirely out of uncertainty. A short, practical policy replaces guesswork with shared standards. The Academy's glossary explains GDPR and copyright.
What to include
| Section | What it covers |
|---|---|
| Purpose and scope | Why the policy exists and who it applies to |
| Approved tools | Which AI tools can be used, and for what |
| Data rules | What information must never be entered |
| Human review | Who checks AI outputs before use |
| Accuracy | Verifying facts, figures and claims |
| Disclosure | When AI use is labelled |
| Copyright and originality | Avoiding infringing outputs |
| Brand voice | Keeping AI content consistent |
| Accountability | Who is responsible for published content |
| Training and updates | How staff learn and how the policy changes |
Approved tools
List tools that are approved, with notes on what each may be used for. Enterprise versions often offer better data protection than free consumer versions. Explain how staff can request new tools and who evaluates them. The Academy's guide on how to evaluate AI tools offers a checklist.
Data rules
Be explicit. Never enter customer personal data, confidential business plans, unreleased product details, passwords or information covered by contracts into public AI tools. Explain the reasons: data may be stored, used for training or exposed. The Academy's guide to UK GDPR for social media explains data protection basics.
Human review
Every AI output used externally should be reviewed by a person. Set expectations for what reviewers check: accuracy, tone, claims, rights and sensitivity. For higher-risk content, such as regulated products or crisis responses, require a second reviewer.
Accuracy and facts
AI tools can state outdated or invented facts. Require verification of statistics, platform specifications, legal points and quotes against reliable sources. The Academy's guide on whether you can trust AI for social media facts explains why.
Disclosure
Decide when AI use is disclosed, such as realistic AI images, synthetic voices or avatars, and follow platform labelling rules. The Academy's guide to AI labels on social media explains requirements.
Copyright and originality
Prohibit prompts that imitate specific artists or recreate trademarked content. Check outputs for resemblance to existing work. Keep human creative input in important assets.
Brand voice
Include links to voice guidelines and approved prompt templates. The Academy's guide to keeping AI content on brand explains how.
What AI should not do
List tasks that stay human, such as replying to complaints, crisis communications and final approval. The Academy's guide on what not to automate on social media offers a starting list.
Agencies and freelancers
Extend the policy to agencies and freelancers through contracts. Ask them to disclose AI use in deliverables and to follow the same data and review rules.
Training
Run short training sessions with practical examples. Show good and poor prompts, review real outputs together and explain the reasoning behind each rule. Policies people understand are policies people follow.
Keeping it current
AI tools change quickly. Review the policy every six months, or sooner when major new tools or rules arrive. Platform facts also change constantly; LinkedIn posts allow 3,000 characters (first ~210 show before 'see more')✓, for instance, and AI tools may quote older limits.
A short policy template
- We use approved AI tools to support, not replace, human judgement.
- We never enter personal or confidential data into public AI tools.
- A person reviews every AI output before publication.
- We verify facts, figures and claims.
- We disclose realistic AI-generated media.
- We keep AI content on brand and original.
- The person publishing is accountable for the content.
Getting buy-in
Policies stick when people see them as helpful rather than restrictive. Involve team members who use AI daily when drafting the policy, include practical tips alongside rules and share examples of time saved by using AI well. Leaders should model the policy in their own use, especially around data and review.
Recording AI use
Ask teams to note which AI tools contributed to major campaign assets. Records help answer questions later about rights, disclosure or accuracy.
Where to keep the policy
Store it somewhere everyone can find, and link it from prompt libraries.
Handling breaches
Explain what happens if someone breaks the policy, such as pasting customer data into a public tool. Focus first on fixing the issue, for example deleting data where possible and assessing whether a data breach report is needed, and then on learning. A blame-free approach encourages people to report mistakes quickly.
A worked example
A retail marketing team discovers staff pasting customer complaint emails into a free chatbot to draft replies. It introduces a two-page AI policy, approves an enterprise AI tool with data protection terms, creates prompt templates for common tasks and runs a one-hour training session. AI use continues to grow, but within clear limits, and complaint replies are reviewed by trained staff.
Common mistakes
- No policy, leaving staff to guess.
- Policies too long or vague to follow.
- Banning AI entirely, pushing use underground.
- Ignoring agencies and freelancers.
- Never updating the policy.
Learning AI governance
The Academy's AI and automation at scale module and team and governance module in the leader level cover AI policies, approvals and risk. The free assessment shows where you would start.
