Rules and ethics
Part of Which rules and ethics govern AI content operations?
Six steps for an AI disclosure policy
A six-step guide to an AI disclosure policy for content operations teams, covering roles, content types, labels, approvals, record-keeping and review.
What to take away
- An AI disclosure policy is a written rule set that says who approves AI-assisted content, what gets labelled, and where the label sits.
- UK-facing content operations fall under UK GDPR and ICO guidance; EU rules may also apply when a campaign falls within their territorial scope.
- Platform terms are separate obligations. They apply even where UK law is silent.
- The decision you are making is not whether to disclose, but who signs off each label before publication.
Who the policy binds
Name every role that touches published content: commissioning editors, freelance writers, sub-editors, and agency partners. A policy that only covers staff leaves contractor output ungoverned. State at the top which territories the policy covers. UK GDPR applies UK-wide; check separately for requirements that vary by territory or arise in other markets.
Who and what the policy binds
- Commissioning editors
- Freelance writers
- Sub-editors
- Agency partners
- England-published content
- Scotland/NIseparate check
- TriggersAI text, synthetic images, cloned voice
State the trigger. Common triggers are text drafted wholly or partly by a generative tool, synthetic images, and cloned voice. Set a threshold so routine spellcheck does not count.
Record the lawful basis for any personal data used in prompts or source material. The ICO's direct marketing guidance explains consent and legitimate interests for promotional messages sent to named people. Keep that reasoning in the policy, not in a separate memo, so the audit trail is one document.
Freelance and agency contracts should require the same standard. Ask for written confirmation at delivery that the piece matches your trigger definitions.
Separate law from platform terms
UK law does not impose a single AI label on editorial content. Consumer protection law still bites if a claim misleads. Advertising rules from the ASA apply to paid promotion.
Law vs platform rules
UK law
- AI label
- No single duty
- Misleading claims
- Consumer law
- Paid promotion
- ASA rules
- EU audiences
- GDPR duties
- Review cycle
- Stable
Platform terms
- AI label
- Channel-specific
- Misleading claims
- Ad policies
- Paid promotion
- LinkedIn terms
- EU audiences
- DSA moderation
- Review cycle
- Changes without notice
Platform rules sit on top. Paid distribution through LinkedIn's marketing tools carries its own ad policies, so check them before scheduling a campaign.
Where content reaches EU audiences, assess applicable EU law, including whether the Digital Services Act (DSA) and EU GDPR apply to the platform and data processing involved. EDPB guidance can help interpret data-protection duties, but it is not a substitute for checking the obligations that apply to your particular campaign. The wider AI content operations rules and compliance guide sets out how the other regimes fit together.
Search and social platforms publish their own disclosure expectations, and they change without notice. Diarise a check of each channel you use so the policy does not drift out of date.
Build the policy in six steps
- List the content types you publisharticles, landing pages, email, social posts, video scripts.
- For each type, state whether a label is required, optional, or prohibited.
- Assign one named approver per type, with a deputy for absence.
- Write the label wording and placement rule, for example a footer line reading "Drafted with AI assistance and edited by our team".
- Set a retention period for prompts, drafts and approval records. Twelve months is a workable default for England-only campaigns.
- Add a review date, at least annual, and put it in the same calendar as your other compliance reviews.
Keep the wording plain. Readers should not need a glossary to understand a label. Test the draft label on two colleagues outside the content team before you publish it. If they hesitate, the wording is doing too much work.
Train and evidence it
Run a short induction for anyone who publishes. Cover the label wording, the approval route, and what to do when a tool's output is unclear. File the induction note with the policy so new starters can read it without waiting for a session.
Keep an approval log. A dated record showing who checked what is the practical evidence if a complaint arrives. Log the tool name and version too, because model behaviour changes between releases.
Review the log quarterly. If approvals cluster on one person, the deputy arrangement is not working.
If suppliers produce content for you, the disclosure duty needs to sit in the contract, not just the style guide. Our piece on AI content operations commercial contracts covers the clauses worth adding.
Common questions
Does England law require an AI label on every article?
No single statute mandates a label on editorial content. Disclosure duties arise from consumer protection law, advertising rules, and platform terms depending on how the content is used. Platform terms often go further than the statute.
Does UK GDPR apply to AI-assisted content?
It applies where you process personal data, for example in source material or marketing lists. The ICO guidance above covers promotional messaging. This is general guidance and individual cases need a qualified adviser.
How often should the policy be reviewed?
At least once a year, and whenever a platform changes its ad or content rules. Put the review date in the document itself.
Who should approve a disclosure label?
One named owner per content type, usually the commissioning editor. Avoid committee approval, which slows publishing without improving accuracy.
