The short answer
The AI Act today asks four things of a manufacturing company that uses artificial intelligence (AI) tools: not to use prohibited AI practices, to take measures for the AI literacy of its staff, to meet the transparency obligations in force since 2 August 2026 and to know whether any system is high-risk. The high-risk obligations have been postponed to 2 December 2027 and, for AI embedded in regulated products, to 2 August 2028.
- The company is almost always a deployer, not a provider: the obligations are lighter.
- Already in force: prohibited AI practices, AI literacy of staff, transparency.
- High-risk obligations postponed: 2 December 2027 for the uses in Annex III, 2 August 2028 for products.
- Take care with human resources: recruiting, evaluating and monitoring workers are high-risk uses.
- In Italy, as one country's example: notice to workers (Law No. 132/2025), criminal and civil liability since 30 September 2026 (Legislative Decree No. 160/2026) and, since 23 October 2026, authorities and fines (Legislative Decree No. 179/2026): the National Cybersecurity Agency (ACN) supervises, and decisions on the employment relationship stay with a person.
Note. Informational content updated as of 12 October 2026, based on the official texts cited below. The national rules described are those of Italy; other EU member states have their own. This is not legal advice: for compliance decisions, consult a professional.
First question: what is your role?
The AI Act, Regulation (EU) 2024/1689, assigns obligations according to role. For a manufacturing company there are three possible roles.
- Deployer. You use AI systems developed by others: an assistant, an agent in the ERP (the company’s management software), vision software for quality control. This is by far the most common case, and the one with the fewest obligations.
- Provider. You develop an AI system, or have one developed, and place it on the market or put it into service under your own name. This is the case, for example, for those who sell their customers a product or a service with AI inside. Take care: you also become a provider by substantially modifying someone else’s high-risk AI system or by changing the intended purpose of a system: anyone who takes a generic assistant and uses it to rank CVs assumes the obligations of the provider of a high-risk AI system.
- Product manufacturer. You build machinery or other regulated products, and AI is a safety component of them. This is a separate chapter, which the 2026 Omnibus changed: we come to it further down.
Anyone who has a system developed for internal use and puts it into service under their own name is also a provider under the Regulation. It matters little as long as the system is not high-risk; it matters a great deal if it is. So the first thing to do is to understand what the system does, not who wrote it.
The updated timeline
In July 2026 the timeline changed. Regulation (EU) 2026/1744, known as the Digital Omnibus on AI, was published in the Official Journal of the European Union on 24 July 2026 and has been in force since 27 July 2026.
| Date | What applies |
|---|---|
| 1 August 2024 | Entry into force of the AI Act |
| 2 February 2025 | Prohibited AI practices and AI literacy |
| 2 August 2025 | Rules on general-purpose AI models, authorities and penalties |
| 27 July 2026 | The Omnibus in force: Article 4 rewritten, high-risk obligations postponed, simplifications |
| 2 August 2026 | General application, including the transparency obligations of Article 50 |
| 23 October 2026 | In Italy: Legislative Decree No. 179/2026, powers of the National Cybersecurity Agency (ACN) and the other authorities, fines, employment decisions reserved for a person |
| 2 December 2026 | New prohibitions on non-consensual intimate content and child sexual abuse material; deadline for marking content generated by systems already on the market before 2 August 2026 |
| 2 December 2027 | Obligations for the high-risk AI systems of Annex III |
| 2 August 2028 | Obligations for high-risk AI linked to the products of Annex I |
High-risk AI systems already on the market before these dates come under the obligations only if they undergo significant changes from then on.
What already applies today
1. Prohibited AI practices
Among others, the Regulation prohibits systems that manipulate people by underhand means, social scoring, biometric categorisation based on sensitive characteristics and, a point that directly concerns companies, emotion recognition in the workplace, except for medical or safety reasons. Software that “measures the attention” or “the mood” of operators is to be avoided.
2. AI literacy of staff
Article 4, in the version in force since 27 July 2026, says that providers and deployers take measures to support the development of AI literacy among their own staff and those who use the systems on their behalf, taking into account knowledge, experience, training and context of use. The text specifies that this does not require them to guarantee any specific level of AI literacy of any individual. The Commission and the member states must support businesses, SMEs in particular, including by publishing practical examples.
In practice: training proportionate to who uses what, internal rules on the use of the tools, and a record of what has been done. The European Commission’s clarifications, with an example of a register, are in the guide AI literacy: what the AI Act requires of companies.
3. Transparency
Article 50 has applied since 2 August 2026. For a company that uses AI there are three practical cases.
- Assistants and chatbots aimed at customers or suppliers: people must know that they are interacting with an AI system, unless it is obvious. The system must be designed this way by the provider; whoever puts it on their own website must check that the notice is there.
- Artificial images, audio and video that look real (deep fakes): whoever distributes them must disclose that they have been generated or manipulated.
- Emotion recognition or biometric categorisation, in the few cases where they are allowed: the people exposed to them must be informed.
Providers of systems that generate content must also mark it in a machine-readable format; for systems already on the market before 2 August 2026 the deadline is 2 December 2026.
Which uses can be high-risk in a manufacturing company
Human resources. Annex III, point 4, classifies as high-risk the systems used to recruit staff, that is, to place targeted job advertisements, filter applications and evaluate candidates, and those used to decide on promotions or terminations, allocate tasks based on behaviour or personal characteristics, and monitor and evaluate the performance and behaviour of workers. Software that ranks CVs or assesses the productivity of individuals falls here.
AI as a safety component of a machine. Here the Omnibus changed course. The Machinery Regulation (EU) 2023/1230, which applies from January 2027, has been moved to Section B of Annex I to the AI Act: the requirements for high-risk AI used as a safety component will be written into the Machinery Regulation itself, through Commission delegated acts that will have to apply by 2 August 2028. For machine builders this means a single frame of reference, the sectoral one, instead of two parallel compliance routes. The Omnibus also clarifies that systems used only for user assistance, performance optimisation, efficiency, automation or quality control are not safety components, unless their failure endangers health and safety.
What is not normally high-risk. Assistants on company documents, agents that read orders and fill in customer and supplier records, translations, preparing quotes, checking invoices, analysing market data, and predictive maintenance and quality control that perform no safety function. For these, the three general obligations seen above apply.
If you use a high-risk AI system: the deployer’s obligations
They will apply from 2 December 2027, but it is worth knowing them before choosing a provider. Article 26 requires anyone who uses a high-risk AI system to:
- use it in accordance with the provider’s instructions for use;
- assign human oversight to people who are competent, trained and have the authority to intervene;
- make sure that the input data under their control is relevant;
- monitor its operation and, in the event of a risk or a serious incident, suspend its use and notify the provider and the authorities;
- keep the logs generated by the system for at least six months;
- inform workers’ representatives and the affected workers before using the system in the workplace;
- cooperate with the authorities.
The national layer: the example of Italy (Law No. 132/2025)
In Italy, Law No. 132 of 23 September 2025 (the country’s AI law), in force since 10 October 2025, sits alongside the European Regulation. For businesses there are two points to know.
- Employment (Article 11). The use of AI in the workplace must be safe, reliable and transparent. The employer must inform the worker of the use of artificial intelligence (AI) in the cases and in the manner set out in Article 1-bis of Legislative Decree No. 152/1997, that is, when using fully automated decision-making or monitoring systems that affect recruitment, management of the employment relationship, allocation of tasks, surveillance or evaluation of workers. The system does not have to be high-risk for this obligation to apply. An employer who fails to inform risks the administrative penalty laid down for information obligations, from EUR 100 to EUR 750 for each month concerned, with higher amounts when more than five or more than ten workers are involved.
- Authorities (Article 20). The national AI authorities are Italy’s National Cybersecurity Agency (ACN), the market surveillance authority with inspection and sanctioning tasks, and the Agency for Digital Italy (AgID), the notifying authority. Powers, procedure and fines are in Legislative Decree No. 179/2026, in force since 23 October 2026: see below.
New since 30 September 2026: Legislative Decree No. 160/2026. Published in the Gazzetta Ufficiale (Italy’s official gazette) on 15 September 2026, it implements the delegation in Law No. 132 on civil and criminal liability. Three points affect businesses.
- A new criminal offence (Article 437-bis of the Italian Criminal Code). It punishes with imprisonment of one to five years anyone who omits the technical safety measures required for high-risk AI systems, or the human oversight measures, when this results in a danger to life or personal safety. It also covers the professional deployer who intentionally omits human oversight.
- Corporate liability. The offence enters Legislative Decree No. 231/2001 with a fine of six hundred to one thousand quotas (the unit in which Italian law sets corporate fines) and disqualification sanctions: it must be taken into account in organisational models.
- Damages claims. When the damage results from a breach of obligations under the AI Act, the causal link between the breach and the damage is presumed, unless proven otherwise.
Since 23 October 2026: Legislative Decree No. 179/2026 on authorities, supervision and penalties. Published in the Gazzetta Ufficiale on 8 October 2026, it completes the Italian framework. For a company that uses AI there are six points.
- Who supervises. The National Cybersecurity Agency (ACN) is the market surveillance authority for AI systems and the single point of contact: it receives complaints, requests documents and information, carries out inspections and can order corrections, withdrawals and suspensions, with the powers of the European regulation on market surveillance. For banks, financial intermediaries and insurers, supervision lies with the Banca d’Italia (the central bank), CONSOB (the securities market regulator) and IVASS (the insurance supervisor). The Garante (Italy’s data protection authority) is a surveillance authority only for certain high-risk uses and keeps its powers over personal data. AgID remains the notifying authority for the bodies that certify high-risk AI systems.
- The fines are those of the AI Act, applied in Italy by the ACN: up to EUR 35 million or 7% of worldwide annual turnover for prohibited AI practices; up to EUR 15 million or 3% for infringing the obligations of providers, importers, distributors and deployers of high-risk AI systems and the transparency obligations; up to EUR 7.5 million or 1% for incorrect, incomplete or misleading information to the authorities. The decree adds a band of up to EUR 1 million or 0.5% for those who do not carry out the fundamental rights impact assessment or do not explain to the person concerned a decision taken with a high-risk AI system (Articles 27 and 86 of the Regulation). For SMEs, microenterprises and innovative start-ups the lower of the amount and the percentage applies.
- For minor infringements the authority may, instead of a fine, order the infringement to be ended within a set time or, if it has already ceased, publish a statement giving the infringement and the name of the party responsible. Anyone who does not comply with the order pays the fine, increased by up to one third. Extracts of the decisions are published on the authority’s website. There is no option to settle the fine by paying a reduced amount.
- AI literacy under Article 4 carries no fine in Italy either. The transparency obligations of Article 50 do.
- On employment the decree goes beyond the AI Act. Final decisions on hiring, on changing or ending the employment relationship, disciplinary measures and performance evaluations cannot be taken by an automated system alone: the decision belongs to a person with effective and independent power, the worker’s consent is not enough, and a dismissal decided in breach of this rule is void. Before these systems are used, the notice required for automated decision-making systems must be given, and the worker can ask for an understandable statement of the reasons for the decision, including the weight the AI had. AI systems that affect the organisation, pace, methods or safety of work enter the workplace risk assessment, with information and training for workers on the specific risks.
- From when. The decree has been in force since 23 October 2026, but the fines follow the AI Act timeline: prohibited AI practices and information to the authorities already now, transparency since 2 August 2026, the obligations of deployers of high-risk AI systems from 2 December 2027. The detailed rules on the ACN’s checks and procedure are left to a separate regulation.
The penalties
Article 99 provides for penalties of up to EUR 35 million or 7% of worldwide annual turnover for prohibited AI practices, up to EUR 15 million or 3% for infringing the obligations of providers, importers, distributors and deployers of high-risk AI systems and the transparency obligations (AI literacy under Article 4 is not among them, and the Italian decree has not added a fine for it) and up to EUR 7.5 million or 1% for incorrect or misleading information to the authorities. For SMEs the lower of the amount and the percentage applies; the Omnibus extended a similar rule to small mid-cap companies for the last two bands. In Italy the fines have been applied by the ACN since 23 October 2026 (Legislative Decree No. 179/2026), which adds a band of up to EUR 1 million or 0.5% for the fundamental rights impact assessment and for the explanation of decisions taken with high-risk AI systems; for minor infringements the authority may limit itself to an order to remedy within a set time.
What to do now: seven steps
- Draw up an inventory of the AI systems in use, including those that people use on their own initiative.
- Classify each system: the company’s role, the use, the risk level.
- Remove or correct the uses that come close to prohibited AI practices, starting with anything that “reads” workers’ emotions or behaviour.
- Organise AI literacy: training proportionate to roles, internal rules on confidential data and permitted tools, a register of what has been done.
- Put transparency in order: a notice on public-facing assistants, a label on realistic artificial content.
- Inform workers when you use fully automated decision-making or monitoring systems that affect the employment relationship, and keep every decision on hiring, evaluations, disciplinary measures and dismissals in the hands of a person: this is what the law requires in Italy. If AI affects the pace and safety of work, update the risk assessment.
- Look at the contracts with your providers: instructions for use, logs, liability, where the data goes. And mark 2 December 2027 in your calendar.
How we take this into account in our projects
In the systems we build, the rules of the AI Act coincide with good design: people in the decision-making loop, an action log, transparency about where the AI is at work, permissions identical to those of the users, and written instructions for those who will use the tool. This is our method, and it is how the CRM with AI agents we use every day works.
We do not develop high-risk AI systems without a specific assessment shared with the client and the client’s advisers.
Sources
- Regulation (EU) 2024/1689 (AI Act), text in the Official Journal of the European Union
- Regulation (EU) 2026/1744 of 8 July 2026 (the Digital Omnibus on AI), OJ L of 24 July 2026
- European Commission, regulatory framework for AI and application timeline
- Italy's Law No. 132 of 23 September 2025, provisions and delegations of power to the Government on artificial intelligence, on Normattiva (in Italian)
- Italy's Legislative Decree No. 160 of 9 September 2026, civil and criminal liability relating to AI, Gazzetta Ufficiale No. 214 of 15 September 2026 (in Italian)
- Italy's Legislative Decree No. 179 of 7 October 2026, powers of the national AI authorities and penalties, Gazzetta Ufficiale No. 234 of 8 October 2026 (in Italian)
- Italy's Legislative Decree No. 179 of 7 October 2026, text on Normattiva (in Italian)
- Regulation (EU) 2023/1230 on machinery
- European Commission, AI literacy: questions and answers
- European Commission, transparency obligations under Article 50
- European Commission AI Act Service Desk, implementation timeline
Checked on 28 October 2026. Rules and deadlines change: always check the official source before you decide.