AI Act transparency obligations for companies: what changes in 2026
AI Act transparency obligations for companies are back in focus following changes that took effect on 2 August 2026. The central issue is not only what the new rules require, but also which role your organisation plays under the regulation: a company may provide, import, distribute or simply use an AI system in its professional activity.
This distinction matters now because the source material indicates that the new directives affect both organisations supplying AI models and those using them. It also notes that some obligations have been postponed, making a clear view of what applies today—and what remains uncertain—especially important.
What the source says changed on 2 August 2026
According to the source material, new AI Act obligations entered into force on 2 August 2026. The text places transparency at the centre of the discussion and says these directives concern companies that provide AI models as well as companies that use AI.
However, the material does not list the individual transparency duties, their exact scope, the documents or notices required, or the enforcement arrangements. It would therefore be inaccurate to treat every possible AI Act requirement as confirmed by this source. The practical conclusion is narrower: organisations need to identify their role and assess the duties connected to that role.
A regulation with more than one affected business role
The source identifies several roles within the regulation. They are not interchangeable. A company’s obligations depend on how it participates in putting an AI model or system on the European market, or how it uses such technology at work.
This is a significant point for internal governance. A business may be accustomed to viewing AI regulation as a matter for technology vendors alone. The source explicitly challenges that assumption: companies do not fall outside the discussion simply because they do not sell an AI product.
Why transparency is a business-wide issue
Transparency responsibilities can involve more than the technical team. The people selecting tools, approving procurement, managing employees’ use of chatbots and placing products on the market may all hold information needed to establish the organisation’s role. The source does not prescribe a particular governance model, but it makes the need for this internal clarification apparent.
Four roles companies need to distinguish
Providers of AI models
Providers are organisations that develop or market an AI model, whether free of charge or paid for. The source gives OpenAI, Anthropic and Mistral AI as examples of providers. For this group, the starting point is the model or product they develop or make available.
Because these businesses are closest to the technology being offered, they should not assume that a commercial arrangement or a free-access model changes the relevance of the role. In the source’s description, both free and paid offerings can fall within it.
Importers and distributors
Importers or distributors place systems developed by actors outside the European Union on the European market. The source offers the example of a French company commercialising a tool from a US or Chinese publisher.
This role is particularly important because it shows that a company does not need to create an AI system to have responsibilities related to its market presence. The relevant question is whether it is making an externally developed system available in the European market, rather than who wrote the underlying technology.
Deployers using AI professionally
Deployers use AI in the course of a professional activity. The source says that a company whose employees use ChatGPT, Claude or another chatbot at work is associated with this role. For many organisations, this is likely to be the most immediate lens through which to examine AI Act transparency obligations for companies.
The role may arise through ordinary workplace use, not only through a large-scale technology project. That means informal adoption deserves attention alongside tools that have gone through a formal purchasing process.
The most common misunderstanding: only sellers are covered
The source identifies a frequent error: assuming that the AI Act concerns only businesses that sell AI and, therefore, only providers. Its message is clear that this framing is too limited.
In practice, every company has a role, according to the source, and most companies in France are deployers. While the source is focused on the French business context, the underlying lesson is broadly useful: an organisation should begin with its actual relationship to AI, rather than with the label it gives itself.
One company can need more than one review
The source describes distinct roles but does not state whether a company can hold multiple roles at the same time. Organisations should therefore avoid drawing unsupported conclusions. A prudent reading is to map each relevant activity separately: developing or marketing a model, bringing an external system to the European market, and using AI in professional operations.
This activity-based review helps avoid a blind spot in which a company looks only at its customer-facing product while overlooking employee use of an external chatbot, or vice versa.
What the source does not settle
The supplied material does not specify how to classify complex supply chains, mixed business models or particular AI use cases. Nor does it define the exact transparency measures that attach to each role. Those limits should be documented rather than concealed when a company assesses its current position.
How to approach compliance today
With some obligations reportedly postponed, companies face two parallel tasks: identify what is already relevant after 2 August 2026 and keep track of obligations that will apply later. The source does not provide the postponed dates or the content of those delayed requirements, so no definitive timetable can be drawn from it.
Still, the role definitions provide a practical starting point for an internal review. The aim is not to claim compliance on the basis of a broad statement, but to establish where the organisation participates in the AI ecosystem and where information is missing.
- List AI models and systems developed, marketed, distributed or used in professional work.
- Assign a preliminary role to each activity: provider, importer or distributor, or deployer.
- Include employee use of workplace chatbots, not only centrally purchased AI tools.
- Record which transparency obligations and postponed measures are not detailed in the available source, so they can be checked against the applicable regulatory materials.
Questions for leadership and operational teams
Leadership can ask whether the organisation supplies an AI model, commercialises a system created outside the European Union, or permits and relies on AI in day-to-day professional activity. Operational teams can then identify the tools and practices that support the answers.
These questions are useful because they counter the idea that regulatory relevance begins only when a company sells an AI product. They also create a clearer basis for discussing responsibilities across product, procurement, legal, technology and workforce functions, without assuming details the source does not provide.
Conclusion
The changes reported from 2 August 2026 make AI Act transparency obligations for companies a current governance issue, not a concern reserved for AI developers. The source’s core warning is that providers, importers, distributors and deployers each need to understand their position.
For most businesses, the first step is to recognise that professional use of AI can place the organisation in the deployer role. From there, a careful inventory of AI activities, a role-by-role assessment and explicit tracking of obligations whose timing or content remains unclear offer a more reliable response than assuming the AI Act applies only to sellers.
