Guide › For data protection officers

Information governance pack: licensing company data for AI training (UK)

If you're a data protection officer or information governance lead and someone in your organisation wants to license its email, chat and documents for AI training, this pack is for you. It's eleven free templates, written for this specific arrangement under UK GDPR as amended by the Data (Use and Access) Act 2025, with the parts only you can supply left for completion.

These are templates, not legal advice. Complete them with your DPO or a qualified adviser. They reflect law and guidance as at 24 October 2026, with the transfer template (05) updated on 8 November 2026. micro1 hasn't endorsed them, and anything they say about how micro1 operates must be confirmed by micro1 in writing before you rely on it.

At a glance

  • Most of the Data (Use and Access) Act's data protection changes took effect on 5 February 2026. DLA Piper, Feb 2026, checked Nov 8, 2026 (opens in a new tab) Two matter most here: Article 8A on compatible reuse, and the new recognised legitimate interests, none of which covers commercial licensing for AI training.
  • The likely lawful basis is ordinary legitimate interests, with a careful balancing test. People don't expect their work emails to train AI models.
  • The released data is usually pseudonymised, not anonymous, while anyone holds the key. The ICO's guidance treats identifiability as depending on whose hands the data is in. Pseudonymisation vs anonymisation explains why that matters.
  • A DPIA is almost certainly needed: large volumes of staff and customer communications, reused in a way people wouldn't expect, then sent to the US.
  • Plan on a transfer agreement. When I checked, micro1 Inc. didn't appear on the US Data Privacy Framework list, so the UK-US data bridge doesn't cover it. Use the IDTA or the UK Addendum, with a transfer risk assessment. Data Privacy Framework list (US Department of Commerce), checked Nov 8, 2026 (opens in a new tab) Sending the data to the US

The templates

Each is a Word document you can edit. Text in yellow brackets needs completing or confirming, and shaded guidance notes explain each section.

No.TemplateWhat it does
00Read me firstHow to use the pack, the order of work and the law it reflects.
01Data protection impact assessmentThe core assessment, following the ICO's sample DPIA structure, with a pre-filled risk register and the questions to put to micro1.
02Legitimate interests assessmentThe purpose, necessity and balancing tests for the likely lawful basis.
03Compatibility assessment (Article 8A)Whether reusing each data source for AI training is compatible with why it was collected.
04Scope and data inventory scheduleWhat's in scope, standard exclusions and the screening checks. Your evidence for data minimisation.
05International transfer and transfer risk assessmentThe safeguard for transfers to the US, and a transfer risk assessment if one is needed. Updated 8 November 2026.
06Contract and data licence checklistTerms to look for in micro1's agreement, and checks on your own client contracts.
07Transparency packPrivacy notice wording, a staff briefing, an objection process and a record of processing entry.
08Pseudonymisation and re-identification assessmentHow identifiable the released data is, who holds the key, and a motivated intruder test.
09Decision recordOptions, conditions before release, residual risks accepted and sign-off.
10Addendum: companies no longer operatingAuthority to sign, retention, reaching former staff and handling requests after closure.

A suggested order of work

  1. Ask micro1 first. Send the questions in the annex to the DPIA and wait for written answers. Much of the rest depends on them, especially who de-identifies the data, where, and who holds the key.
  2. Agree a narrow first scope (04) and check your client contracts for confidentiality or data-use limits (06, part C).
  3. Complete the compatibility and legitimate interests assessments (03 and 02).
  4. Complete the DPIA and the transfer assessment (01 and 05).
  5. Review micro1's contract against the checklist (06).
  6. Tell people. Update notices, brief staff and run an objection window before export (07).
  7. Test samples of the de-identified output (08).
  8. Record the decision (09). If high risk remains after the measures, consult the ICO before going ahead.

Where the hard questions are

Free text

Names, signatures, nicknames and project names turn up inside email and chat. Ask how they're found and replaced, and test samples. How micro1 says it de-identifies data

Expectations

Staff didn't write their emails expecting them to train AI. The balancing test, the notices and the objection window carry most of the weight.

Client confidentiality

Contracts with clients may restrict how their information is used, whatever data protection law allows. Most organisations leave client files out of scope.

Special category data

Health, union and similar information turns up in HR threads and sickness emails. Screen for it and exclude it.

Sending the data to the US

Where the names are removed decides how much crosses the Atlantic. If micro1 de-identifies the data on its own systems, fully identifiable data leaves the UK first, and the key that links stand-ins back to real people stays in the US.

StageWherePersonal data?What you need
1. ExportRaw email, chat and documents go from the UK to micro1Yes, fully identifiableA transfer safeguard, and processor terms for the de-identification work (template 06, part A)
2. De-identificationmicro1 replaces names and holds the key and, for a time, the raw copyYes, for micro1, because it holds the keyWritten limits on how long the raw copy and the key are kept, security, and no other use
3. LicensingThe pseudonymised copy goes to labs in the US and allied countriesPossibly not, for a lab that can't realistically re-identify anyoneContract terms that ban re-identification and onward sharing to anyone who could

The safeguard. The UK-US data bridge only covers US companies certified to the Data Privacy Framework and its UK Extension. When I checked, micro1 Inc. didn't appear on that list. Data Privacy Framework list (US Department of Commerce), checked Nov 8, 2026 (opens in a new tab) So plan on the International Data Transfer Agreement (IDTA) or the UK Addendum to the EU standard clauses, with a transfer risk assessment under the "not materially lower" test that has applied since 5 February 2026.

Why a fallback matters anyway. On 29 June 2026, the US Supreme Court ruled in Trump v Slaughter that the President can remove commissioners of the Federal Trade Commission, the body that enforces the framework, at will. NOYB, the group behind earlier successful challenges, has said it will challenge the EU-US framework at the EU's top court. Society for Computers and Law, Jul 2026, checked Nov 8, 2026 (opens in a new tab) The UK data bridge is a separate UK regulation and still in force, but the government can revoke it. Even with a certified recipient, have the IDTA agreed in advance.

The lower-risk option. Ask whether de-identification can run in the UK or EU, or in your own environment, with the key kept in the UK. Then only pseudonymised data crosses the Atlantic. For you it's still personal data, so treat it as a transfer, but the risk assessment becomes much simpler.

Questions to put to micro1 in writing:

  1. Are you certified to the Data Privacy Framework and the UK Extension? If not, will you sign the IDTA or the UK Addendum?
  2. Where does de-identification run, and on whose infrastructure?
  3. Who holds the key, and when is it destroyed?
  4. How long is the raw copy kept, and how is deletion confirmed?
  5. Which labs receive the data, in which countries, and on what terms against re-identification and onward sharing?

Common questions from DPOs

Do we need a DPIA to license company data for AI training?

Almost certainly. It reuses large volumes of staff and customer communications for a purpose people wouldn't expect, involves new technology and sends data to the US. The pack's DPIA follows the ICO's sample structure and pre-fills the likely risks.

What lawful basis applies to licensing data for AI training?

Usually legitimate interests, with a legitimate interests assessment. The new recognised legitimate interests cover purposes such as security, crime prevention and safeguarding, not commercial licensing. Consent is rarely practical for years of past communications.

Is reusing work email for AI training a compatible purpose?

It has to be assessed source by source under Article 8A, looking at the link between purposes, the context, the nature of the data, the consequences and the safeguards. De-identification and an objection window help, but they don't decide it on their own.

Is de-identified data still personal data?

For your organisation, usually yes, because you hold the originals. For a recipient with no realistic means of re-identifying anyone, it may not be. Pseudonymisation vs anonymisation covers the ICO's approach and the 2025 EU court ruling.

Can we transfer the data to a US company?

Yes, with a valid safeguard. A recipient certified to the UK Extension to the EU-US Data Privacy Framework is covered by the UK-US data bridge, but micro1 Inc. didn't appear on that list when I checked. Use the IDTA or the UK Addendum, with a transfer risk assessment, and keep it as a fallback even for certified recipients. Template 05 covers both.

Is there a US or Canadian version?

Not yet. The pack is written for UK law. US and Canadian privacy law uses different tests, so the templates shouldn't be used there without adaptation.

About these templates

I wrote them from a background of 22 years delivering data and digital projects into the NHS. I'm not a lawyer. I'm also a referral partner for micro1: if a company signs up through my introduction, micro1 may pay me a referral fee. The templates are free to use and adapt, with or without my name on them. If something in them is wrong or out of date, please tell me.

Sean Key

About the author

Sean Key has been a freelance consultant for 22 years, mainly delivering data and digital projects into the NHS. He runs Applied Clinical Judgement, which connects professionals with AI training work, and helps established companies weigh up micro1's Company Data Partnerships program.

He isn't a lawyer and isn't employed by micro1. If your company signs up through his introduction, micro1 may pay him a referral fee at no cost to you.

Sean Key on LinkedIn (opens in a new tab) · Book a call (opens in a new tab)

Find out if your company qualifies

Apply through my referral link. micro1's form takes about 10 minutes, and its team checks whether your data fits. There's no cost and no obligation.

Register with micro1 (opens in a new tab)micro1's registration pagemicro1 will ask you to enter your work email to register.
Still not sure?

Ask me first. I'll tell you plainly whether your company looks like a fit.

Send me a message Book a call (opens in a new tab)