AI Act 2026: What Changes for Businesses and Which Obligations Already Apply

A virtual assistant responding to customers, software supporting recruitment processes, a system generating content: the use of artificial intelligence can involve very different responsibilities. Following the provisions that became applicable in 2026 and the changes introduced by the AI Omnibus, businesses need to distinguish between obligations that already apply and those scheduled for the coming years. The AI Act (Regulation (EU) 2024/1689) does not impose the same obligations on every organisation.
AI Act 2026: Changes Introduced by the AI Omnibus
On 27 July 2026, Regulation (EU) 2026/1744, known as the AI Omnibus, entered into force, introducing amendments and simplifications to the AI Act. Its aim is to facilitate the implementation of the rules by reducing certain administrative burdens on businesses and changing the deadlines for high-risk AI systems. Key developments include extending certain simplification measures to small mid-cap companies, changes to the registration of specific systems in the EU database, and greater opportunities for testing in controlled environments.
The rules on AI literacy have also changed: organisations concerned must take measures to develop the skills of relevant personnel, without being required to ensure that each individual achieves a specific level of proficiency.
Which AI Act Obligations Already Apply to Businesses?
The AI Act provides for the gradual application of its provisions. To understand which obligations currently affect businesses, it is useful to review the main implementation dates.
Since 2 February 2025, certain uses of artificial intelligence considered particularly harmful to people's rights have been prohibited. For example, using AI systems to infer employees' emotions in the workplace is prohibited, subject to specific exceptions for medical or safety reasons. The provisions on AI literacy have also applied since that date and were subsequently updated in 2026. Providers and deployers of AI systems must take measures to support the development of AI literacy among their staff and other people operating or using AI systems on their behalf, taking account of their knowledge and the tasks they perform.
Since 2 August 2025, specific obligations have applied to providers of general-purpose AI models, meaning models designed to perform a wide range of tasks and capable of being used in different applications. These provisions cover, among other things, technical documentation, information for organisations integrating the models into other systems, compliance with copyright law, and the publication of a summary of the content used for training. These obligations do not automatically apply to every business that uses a chatbot. Furthermore, models placed on the market before 2 August 2025 are subject to a transitional period, with a compliance deadline of 2 August 2027.
Since 2 August 2026, the new transparency provisions for AI systems have become applicable. The European Commission's powers to oversee compliance with the obligations imposed on providers of general-purpose AI models have also come into effect.
Transparency in the Use of Artificial Intelligence
Article 50 introduces different obligations depending on the company's role and the type of application. Providers of systems designed to interact directly with people must ensure that users are informed when they are interacting with AI, unless this is obvious from the circumstances. Providers of generative AI systems must also, where required, implement machine-readable markings that make it possible to identify artificially generated content. Organisations deploying AI systems professionally must disclose certain deepfakes and, under specific conditions, AI-generated text published to inform the public on matters of public interest.
For the latter, an exception applies when the content has undergone human review or editorial control and a natural or legal person assumes editorial responsibility for its publication. This does not mean, therefore, that every AI-generated text must automatically carry a visible label.
Developing or Using AI: Different Responsibilities
A fundamental distinction under the AI Act concerns providers and deployers. A company that develops and markets an AI system under its own brand may qualify as a provider, while an organisation using the system under its own authority in its business processes is generally considered a deployer. For example, using an AI assistant for customer support is not equivalent to developing a system designed to assess job applicants. The risks, information requirements and, where applicable, control measures differ. A company using a third-party system may also assume additional responsibilities if it substantially modifies the system or changes its intended purpose, under the conditions set out in the Regulation.
How Businesses Can Prepare for AI Act Obligations
The first step is to identify where AI is being used, including through services purchased from external providers. For each application, businesses need to establish its purpose, the data being processed, the parties responsible and its potential risk classification. Only then can they identify the obligations that actually apply and the relevant deadlines. It is also necessary to review the information provided to users, human oversight measures, data protection arrangements and the skills of those operating the systems. More complex projects also require activity traceability, controls over AI-generated outputs and documented procedures for managing issues.
Compliance with the AI Act is not simply a matter of purchasing a particular tool: it involves technology, processes and people.
From Compliance to AI Governance
Understanding the deadlines is necessary, but not sufficient. For businesses, the challenge is to maintain control over AI as it becomes increasingly integrated into operational processes. S2E addresses these challenges through expertise and solutions dedicated to AI Governance, including Galene.AI and Generative Shield, designed to support interaction monitoring, policy management and risk reduction. These tools can support an organisation's compliance efforts, without replacing the assessment of its specific regulatory obligations.
FAQ on the AI Act 2026
Does the AI Act Also Apply to Businesses Using Third-Party AI Tools?
Yes. Organisations that professionally deploy AI systems may also be subject to obligations, depending on their role, the intended use and the system's risk classification.
From 2026, Must All AI-Generated Content be Labelled?
No. Article 50 establishes marking or disclosure requirements in specific circumstances, with certain exceptions. It is important to distinguish between system providers and businesses using AI-generated outputs.
When Do the Obligations for High-Risk AI Systems Become Applicable?
For systems covered by Annex III, from 2 December 2027; for high-risk AI systems integrated into regulated products listed in Annex I, from 2 August 2028, subject to the applicable transitional provisions.
Does the AI Act Require All Employees to Undergo AI Training?
The Regulation requires providers and deployers to take appropriate measures to support the AI literacy of people operating AI systems on their behalf. However, it does not prescribe the same training course or a fixed level of proficiency for everyone.

Riempi il Form sottostante per poter lasciare i tuoi commenti