---
title: "AI Act: the obligations for a company that uses AI"
description: "Prohibited practices, AI literacy, transparency and high-risk systems: the AI Act obligations for those who use AI, with the dates after the 2026 revision."
url: https://www.joinx.it/en/resources/ai-act-what-it-asks-of-companies
---

# AI Act: the obligations for a company that uses AI

Prohibited practices, AI literacy, transparency and high-risk systems: the AI Act obligations for those who use AI, with the dates after the 2026 revision.

JoinX · Last updated 18 November 2026

The AI Act, Regulation (EU) 2024/1689, has been in force since 1 August 2024 and applies in stages. In July 2026, Regulation (EU) 2026/1744, known as the Digital Omnibus on AI, amended parts of it and moved some deadlines. In this article we summarise what it asks today of a company that uses AI systems without developing them for the market, with the dates in force as of November 2026. It is a summary to help you find your bearings, while how the law applies to your case is for your legal and compliance functions to establish.

## Provider or deployer

The regulation distinguishes those who develop an AI system and place it on the market, the provider, from those who use it in the course of their own business, the deployer. Most companies are deployers, because they use a licensed assistant, a model from a major vendor or an agent built for one of their processes. The line is not always clear-cut, though. In the case of high-risk systems, whoever substantially modifies one, or puts it on the market under their own name, takes on the provider's obligations.

## What already applies

Since 2 February 2025 the practices listed in Article 5 have been prohibited, including social scoring of people and emotion recognition in the workplace, except for medical or safety reasons. From 2 December 2026 the revision adds to the list systems that generate realistic intimate images of identifiable people without their consent, or child sexual abuse material. For a company, the first check is to verify that no tool in use falls within this list, including those bought by a single function.

Since the same date, Article 4 has asked providers and deployers to attend to the AI literacy of the staff who use these systems. The wording changed with the 2026 revision, and the article now asks for measures to support the development of these skills, specifying that there is no need to guarantee a set level for each person. The obligation remains, so it is worth being able to show what has been done, for which roles and with what outcome.

Since 2 August 2026 the regulation has applied in general, including the transparency obligations of Article 50. For a company, there are two common cases. An assistant that converses with customers or employees must make it clear that they are talking to an AI system. Anyone who uses a system to generate or alter images, audio or video of real people, places or events that could appear authentic, known as deepfakes, must disclose that they have been artificially generated or manipulated.

## High-risk systems

Annex III lists the uses considered high-risk. 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, allocate tasks or evaluate workers' performance.

For these systems the 2026 revision moved the application of the obligations from 2 August 2026 to 2 December 2027. High-risk systems that are part of products already covered by European safety legislation, such as medical devices, instead follow the deadline of 2 August 2028.

Article 26 says what anyone using a high-risk system must do. They must use it in accordance with the provider's instructions and assign its oversight to people with the necessary competence, training and authority. They must check that the input data under their control is relevant, monitor the operation of the system, keep the logs the system generates for at least six months and inform workers and their representatives before using it in the workplace. Banks and insurers that use systems for creditworthiness or for life and health premiums must also carry out a fundamental rights impact assessment, as 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.

## Where to start

The first step is a complete list of the AI systems in use, which often nobody has. An email assistant or a service that summarises meeting minutes can enter the company through one function without compliance knowing. With [Mapping](https://www.joinx.it/en/solutions/mapping) we take an inventory of them 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](https://www.joinx.it/en/technology/governance) we collect the evidence as the work happens, that is, who approved, with which data and under which rules. AI literacy is handled through [Adoption](https://www.joinx.it/en/solutions/adoption), with role-based paths for which we document attendance, content and results.

The documentation we produce has a limit: it 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, instead of presenting them with the project once the work is done. The other commitments we put in writing are on the [About us](https://www.joinx.it/en/about-us#fiducia) page.

## Sources

The text of the AI Act, Regulation (EU) 2024/1689, is published on [EUR-Lex](https://eur-lex.europa.eu/eli/reg/2024/1689/oj). The amendments cited here come from [Regulation (EU) 2026/1744](https://eur-lex.europa.eu/eli/reg/2026/1744/oj), published in the Official Journal of the European Union on 24 July 2026 and in force since 27 July 2026. We update this article when the application timetable changes.

[All articles](https://www.joinx.it/en/resources#articoli)

## Know which AI systems you already have in use

An inventory of the systems in use is often the first useful piece of work, even before a new project.

[Contact us](https://www.joinx.it/en/contact?tema=governance)
