AI Act: the checklist for those who use AI
The AI Act checks for a company that uses AI: inventory, prohibited practices, AI literacy, transparency and high risk from 2 December 2027.
JoinX · Last updated
This checklist is designed for a company that uses AI systems in the course of its business without developing them for the market, that is, for what the AI Act calls a deployer. It follows the dates in force as of November 2026, after the amendments introduced in July 2026 by the Digital Omnibus on AI, and accompanies the article AI Act: the obligations for a company that uses AI, where the obligations are explained in full. It is meant to help you find your bearings and prepare the work, while how the law applies to your case is for your legal and compliance functions to establish.
The list of systems in use
All the later checks depend on this one, because you cannot assess a system nobody knows exists. An email assistant or a service that summarises meeting minutes often enters the company through a single function, without compliance knowing.
- There is a list of the AI systems in use, including licensed assistants and tools bought by a single function.
- For each system the purpose, the supplier, the internal owner and the data it uses are stated.
- For each system it is established whether the company simply uses it, or has substantially modified it or offers it under its own name, cases in which, for high-risk systems, it takes on the provider's obligations.
Prohibited practices, Article 5
The prohibitions have applied since 2 February 2025, and from 2 December 2026 the list gets longer.
- No system in use assigns a social score to people.
- No system recognises emotions in the workplace, except for medical or safety reasons.
- No system generates realistic intimate images of identifiable people without their consent, or child sexual abuse material, prohibitions added by the 2026 revision.
- The check also covers tools bought by a single function.
AI literacy of staff, Article 4
Since 2 February 2025, Article 4 has asked those who use AI systems to attend to the skills of their staff. After the 2026 revision it asks for measures to support AI literacy, without requiring a set level to be guaranteed for each person, but the obligation remains.
- The roles that use AI systems or check their work have been identified.
- For each role there are training measures suited to the use it makes of them.
- Attendance, content and outcomes are documented, so that what has been done can be shown.
With Adoption we organise role-based paths and document their attendance, content and results.
Transparency, Article 50
Since 2 August 2026 the regulation has applied in general, including the transparency obligations.
- An assistant that converses with customers or employees makes it clear that they are talking to an AI system.
- Images, audio and video of real people, places or events generated or altered with AI, which could appear authentic, are disclosed as such.
- Texts generated with AI and published to inform the public on matters of public interest say so, unless a person has reviewed them and holds editorial responsibility for them.
High-risk systems, from 2 December 2027
Annex III lists the high-risk uses. In the industries we work in, the most frequent are the assessment of the creditworthiness of natural persons, risk assessment and pricing in life and health insurance, and systems used to recruit staff or evaluate their performance. For these uses the 2026 revision moved the obligations to 2 December 2027, while systems that are part of products already covered by European safety legislation, such as medical devices, follow the deadline of 2 August 2028. For every high-risk system in use, Article 26 asks you to check the following.
- The system is used in accordance with the provider's instructions.
- Oversight is assigned to people with the necessary competence, training and authority.
- The input data under the company's control is relevant to the purpose of the system.
- Someone monitors the operation of the system and knows when to inform the provider and, where required, the authority, for example in the event of a risk or a serious incident.
- The logs generated by the system are kept for at least six months.
- Workers and their representatives are informed before the system is used in the workplace.
- The people about whom the system helps to take decisions know that a high-risk system is in use.
- Banks and insurers that use systems for creditworthiness or for life and health premiums have carried out the fundamental rights impact assessment provided for in Article 27.
The postponement buys time, but the work needed to be ready is long, because it starts with knowing which systems are in use and who is accountable for them.
What we do, and the limit
With Mapping we take an inventory of the AI systems in use and enter them in a register, each with an owner and a risk level. For what we build, we classify every system under the AI Act before development, and with Governance we collect the evidence as the work happens, that is, who approved, with which data and under which rules. The documentation we produce serves to demonstrate how a system is governed, but on its own it does not certify compliance, which remains the company's responsibility. That is why we involve those responsible for risk, compliance and data protection from the start. The other commitments we put in writing are on the About us page.
Sources
The text of the AI Act, Regulation (EU) 2024/1689, is published on EUR-Lex, and the amendments cited here come from Regulation (EU) 2026/1744, published in the Official Journal of the European Union on 24 July 2026. We update this checklist when the application timetable changes.
The checklist as a PDF
The checks in this guide, section by section, to fill in with your legal and compliance functions. We ask for your contact details, then the download starts straight away.