Do our own AI apps make us a provider under the AI Act?
Usually not in the sense of the high-risk obligations. If you build apps on an AI platform for internal tasks, such as drafts, summaries or questions on documents, you use AI as a deployer. You become the provider of a high-risk system under Art. 25(1)(c) if you build an app for a purpose listed in Annex III, such as assessing job applications. For public-facing chatbots, provider obligations under Art. 50 can arise.
Who is a provider, who is a deployer?
- Provider (Art. 3(3)) is anyone who develops an AI system or has it developed and places it on the market or puts it into service under their own name, whether for payment or free of charge. Putting it into service for one’s own use also counts (Art. 3(11)).
- Deployer (Art. 3(4)) is anyone who uses an AI system under their own authority in a professional capacity.
You use an AI platform with a chat assistant as a deployer. It is an open question when an app you build yourself becomes an AI system of its own. For the obligations, however, two questions matter most: does the app pursue a high-risk purpose, and does it interact with people outside your organisation?
When do you become the provider of a high-risk system?
Art. 25(1) names three cases in which a deployer or another third party is considered the provider of a high-risk system:
| Point | Case | Example |
|---|---|---|
| a | You put your name or trademark on a high-risk system that is already on the market | a purchased recruitment tool appears as your product |
| b | You make a substantial modification to a high-risk system, and it remains high-risk | your own changes to a triage system |
| c | You change the purpose of a system that is not high-risk in such a way that it becomes high-risk, including general-purpose AI | you build an app on a general assistant that assesses job applications or decides on benefits |
Case c is the most important one for platforms. An app that pre-sorts job applications, assesses benefit claims or classifies emergency calls can make your organisation the provider, with all obligations under Art. 16. Check beforehand whether an exception under Art. 6(3) applies, for example because the app only prepares a decision. See Identifying high-risk AI.
The original provider is then no longer considered the provider of that system. Under the newly worded Art. 25(2), however, it must cooperate: technical documentation, known limitations and failure patterns, targeted technical access. This does not apply if it has expressly specified that its system must not be changed into a high-risk system. Ask your provider whether its contract contains such a specification.
What applies to public-facing chatbots?
If you build an app that citizens, patients or customers write to, and make it available under your name, you may be a provider within the meaning of Art. 50(1): people must be able to recognise that they are talking to an AI. The Commission’s guidelines assign the obligations along the supply chain. Details in the guide Transparency obligations under Art. 50.
And for medical apps?
If an app has a medical intended purpose, such as a diagnostic suggestion, it may be a medical device. You are then the manufacturer under the Medical Device Regulation, and the AI Act treats the app as high-risk via Annex I (from 2 August 2028). More on this: Is AI a medical device?
How does this work with basebox?
basebox is a platform, not a high-risk system; which obligations you have depends on how you use it. In basebox, an app consists of an instruction, documents and connections to your systems, and it is created without programming. Your staff can only use what your administrator has enabled: apps, connections to systems and AI models. Whoever builds an app shares it with groups or people, to use or to edit. Your administrator sets usage policies and system defaults centrally; they apply to all apps. The audit log records who used which assistant and which connector, and the developer, version and licence of each model are listed in the model registry. basebox is not built to assess people or to carry out profiling. More under Features and Security.
Not legal advice: As of 11 October 2026. When a configured app counts as an AI system of its own has not been settled by case law. Have planned uses in high-risk areas legally reviewed in advance.
Frequently asked questions
Is an app with an instruction and documents already an AI system of its own?
This has not been conclusively settled in law. What matters in practice is the purpose: as long as the app does not pursue a purpose from Annex III and no one outside your organisation talks to it, the provider role brings hardly any additional obligations.
What do we have to do as the provider of a high-risk system?
Meet the obligations under Art. 16: among others risk management, data governance, technical documentation, logging, instructions for use, human oversight, quality management, conformity assessment, CE marking and registration. For most organisations, this is a project of its own.
How do we prevent someone in our organisation from building a high-risk app?
With clear rules: which purposes are allowed, who may share apps with others, and a review before sharing. Set out the rules in your AI policy and train the people who build apps (Art. 4).
Sources
This article was created with the help of AI, based on the sources listed. Responsible for the content: basebox GmbH. Not legal advice, no liability. Research as of 11 October 2026.