Skip to content

AI Act8 November 2026

Company AI use policy: a template for SMEs and what the law says

Xora Intelligence10 min read
An employee pins the rules for using AI on the noticeboard while a colleague reads them

A company AI use policy is a short document that says which AI tools your team can use, with which accounts, what data never goes in and who checks what comes out. No law requires it under that name, but it helps you comply with the LOPDGDD, the GDPR and the EU AI Act.

Below you have a one-page template to copy and fill in, the rules that back it up (with a link to each one) and how to put it into practice so it doesn't end up as a forgotten PDF. We checked everything on 8 November 2026.

What is a company AI use policy?

It is the internal rule that governs how artificial intelligence is used in your business. It is not a technical manual or an ethics code full of principles. It is a sheet that any employee can understand in five minutes and that answers four questions: which tools are authorised, what you can do with them, what is forbidden and who to tell if something goes wrong.

You need it because AI is already inside your company, whether anyone approved it or not. Someone translates an email with a free chatbot, someone else summarises a contract and another person asks an assistant to draft a complaint. If everyone does this with a personal account, the company does not know what data has left or where it went. This is called shadow AI, and we explain it in detail in the guide to security when using AI in your company.

Banning everything rarely works: people carry on using it from their phones. It works better to provide an authorised tool and clear rules.

Is an AI use policy compulsory?

No Spanish or European law says "you shall have an AI use policy". But three sets of rules push you to have one in writing, and a short policy is the cheapest way to comply with them.

  • The LOPDGDD. Article 87.3 of Organic Law 3/2018 requires the employer to set criteria for the use of digital devices, with the involvement of workers' representatives, and to inform staff of those criteria. If your team uses AI from company computers or phones, the AI rules fit in there.
  • The EU AI Act. Its Article 4 requires anyone who uses AI in their activity to take AI literacy measures for their staff from 2 February 2025. After the Digital Omnibus, in force since 27 July 2026, the obligation was simplified but it still exists. It is explained in how to train your team in AI.
  • The GDPR. If an employee pastes customer data into a tool, the company is still responsible for that processing. We cover it in AI and the GDPR with customer data.

There is a public model you can look at. The Spanish Data Protection Agency published its internal policy on the use of generative AI on 27 November 2025. It only applies to the AEPD itself, but its rules for staff work as a guide for any company.

What should an AI use policy include?

Eight sections are enough for an SME. If one does not apply to you, drop it. What matters is that it fits on one page.

Section What it answers Example for an SME
Purpose and scope Who it applies to All staff, including trainees and collaborators
Authorised tools Which AI you can use and with which account One assistant on a company account; nothing on a personal account
Permitted uses What it is for Drafting, summarising your own texts, translating
Forbidden uses What it is not for Deciding about people, replying to customers without review
Data that stays out The red line ID numbers, health data, payslips, passwords, customer contracts
Human review Who checks before it goes out Whoever sends anything to a customer reviews it first
Alerts and errors Who to tell A named person in charge and a specific channel
Training and review How it is kept alive A session on joining and a review of the policy every six months

The AEPD applies almost the same to its staff: only people who have been trained use generative AI, no tool outside its inventory is used, the person using it reviews everything generated and results are checked against reliable sources. You can read this in the "Policy of use for staff" section of its official document.

AI use policy template for an SME

Copy it, change what is in square brackets and have each person sign it or read it. It is designed for a small team that uses AI assistants to write, summarise and translate.

Artificial intelligence use policy of [Company name] Version [1] · Approved on [date] · Person in charge: [name and position]

  1. Who it applies to. Everyone who works for [company], including trainees and collaborators, when they use AI for company tasks.
  2. Authorised tools. Only [tool or tools], on the company account. AI is not used for work on personal accounts. To add a new tool, ask [person in charge].
  3. Permitted uses. Drafting, summarising your own documents, translating, preparing ideas and proofreading texts.
  4. Forbidden uses. Making decisions about people (hiring, dismissing, granting or refusing a service) without a person deciding. Sending a customer something generated by AI without reviewing it. Impersonating another person with a generated voice or image.
  5. Data that is never entered. Health data, ID numbers, bank account numbers, payslips, passwords, customer contracts or documents and any personal data that is not essential. If in doubt, do not enter it and ask.
  6. Review. The person who uses the AI is responsible for what comes out. Anything that goes to a customer, the website or a public body is reviewed by a person before it is sent.
  7. Transparency. If a customer talks to a chatbot or an assistant of the company, they are told it is an AI.
  8. Errors and incidents. If data that should not have been entered was entered, or the AI gave a serious answer, tell [person in charge] the same day through [channel].
  9. Training. Nobody uses these tools without having attended the internal training session. A record is kept of who has received it and when.
  10. Review of the policy. It is reviewed every six months or whenever a new tool is added.

I have read and understood this policy: [name, signature and date]

If you have legal representation of the workers, show it to them before approving it. We explain this below.

What data can you never put into ChatGPT or another AI?

It depends on the tool and the contract, but one rule works for almost any SME: personal data of third parties and confidential information do not go into a tool the company has not authorised. With an authorised tool, only what is essential goes in.

In practice, the blacklist in the template covers what does the most damage if it leaks:

  • Special categories under the GDPR, such as health data. A clinic or a pharmacy should not paste a medical history into a chat.
  • Identifiers and financial data: ID numbers, bank accounts, payslips.
  • Credentials: passwords, access keys, tokens.
  • Customer documents and contracts, even without personal data, if they could harm the company or the customer.

The AEPD goes further in its own policy: it asks you to minimise even sensitive information that is not personal, such as internal documents or references that could harm the image of the institution or of third parties. For an SME, the equivalent is not pasting a tender offer or the price list agreed with a supplier.

Before authorising a tool, read its terms: whether it uses what you type to train its models, where it stores the data and whether you can delete it. In the guide to AI and data protection you have the questions worth asking.

Do you have to involve the workers or their representatives?

Yes, if you have them. Article 87.3 of the LOPDGDD says that in drawing up the criteria for the use of digital devices "workers' representatives must take part". If your company has a works council or staff delegates, the AI policy should go through them before it is approved. And in any case, staff must be informed.

Two more rules are worth keeping at hand:

  • Workers' Statute. Article 64.4.d gives the works council the right to be informed of the parameters, rules and instructions of the algorithms or AI systems that affect decisions on working conditions, access to and keeping of employment, including profiling.
  • EU AI Act. If one day you use a high-risk AI system at work, such as an automatic CV filter, Article 26.7 requires you to inform the representatives and the workers affected beforehand. After the Digital Omnibus, the high-risk rules in Annex III apply from 2 December 2027.

If AI is only used in your company to write and translate, none of this asks you for a complicated procedure. It asks you to inform and listen.

How do you put it into practice so it doesn't stay on paper?

A policy nobody has read protects nobody. These six steps work in a small team:

  1. Take stock. Ask, without punishing anyone, which AI tools each person uses and what for.
  2. Choose an official tool and set it up on a company account. If you offer a good alternative, shadow AI drops on its own.
  3. Fill in the template with your tools, your sensitive data and a named person in charge.
  4. Show it to the workers' representatives, if there are any, before approving it.
  5. Run a short training session with examples from your real work: what can be pasted, what cannot and how to review what comes out. Keep a record of who attended and when.
  6. Review it every six months, or sooner if a new tool comes in or the law changes.

A typical case in the Campo de Gibraltar

In the Campo de Gibraltar many people work in two languages: customers in Gibraltar, companies that deal with the port of Algeciras, or accommodation in Sotogrande and Tarifa with foreign guests. Translating emails with AI is one of the most common uses, and also one of the riskiest, because the email usually carries names, addresses or order details.

The policy solves this with two rules: translate only with the authorised tool and remove the personal data that is not needed to translate. It is a small measure that avoids the most common mistake.

Frequently asked questions

Is it compulsory to have an AI use policy in the company? Not under that name. But the LOPDGDD requires the employer to set criteria for the use of digital devices and inform staff, the EU AI Act asks for literacy measures from February 2025 and the GDPR makes you responsible for the data your team enters into an AI. A written policy is the simplest way to cover all three.

Can I ban ChatGPT in my company? You can, just as you can set which programs are used on work computers. The problem is that a total ban usually pushes use onto personal phones, where you control nothing. It works better to authorise a tool on a company account, say what data stays out and train the team to review what comes out.

Does each worker have to sign the AI use policy? No rule requires a signature. What the LOPDGDD does require is that workers are informed of the criteria for use. Asking each person to sign or tick that they have read it, with a date, is the simplest way to show that you informed them and to keep a record of the training under Article 4 of the AI Act.

How often should the AI use policy be reviewed? The law sets no deadline. A review every six months is reasonable, because tools change fast. Review it also when you add a new tool, when you put a chatbot in front of customers or when the rules change, as happened with the Digital Omnibus in July 2026.

Does the same policy work for a self-employed person with no employees? Yes, in a short version. A self-employed person has nobody to inform, but is still responsible for their customers' data. It is enough to decide which tool to use, with which account, what data never goes in, and to review everything they send. Writing it down in half a page helps you not skip it when you are in a hurry.

What we do at Xora

Xora Intelligence, an AI consultancy in Los Barrios (Cádiz), does not sell templates or certificates. What we do is a company audit to see which AI your team already uses, what data is at stake and what is worth automating, and we help you put in writing the rules that fit the way you work. It is not legal advice: if your case involves health data or decisions about people, we will recommend that you also involve your own adviser.

If you want us to look at it with you, write to us at help@xoraagencia.com or through contact.

Found it useful? Share it with someone who needs it

Xora IntelligenceAI consultancy and custom software in Los Barrios (Cádiz). We write about what we build for businesses in the province. If you have a question about your case, write to us.

Keep readingRelated articles

The whole blog

Bay of Gibraltar · photo by Cabeza2000 (CC BY-SA 4.0)

Shall we talkabout your business?

Tell us which tasks take up most of your time. We will tell you whether they can be automated and how we would do it. If we can't help, we will tell you that too.