app-entwickeln-lassen.de LogoApp entwickeln lassen

Legal, Privacy & Security

EU AI Act for Apps: What Providers of AI Features Must Know

app-entwickeln-lassen.de editorial team 6 min read
Contents
  1. The EU AI Act in brief: timeline and status in 2026
  2. The four risk levels and where typical app features land
  3. Provider or deployer? Your role determines your duties
  4. Transparency duties: what chatbots and generative features must do
  5. Checklist for AI features in your app
  6. Penalties, risks and consequences for your app project
  7. Conclusion
  8. Frequently Asked Questions

The EU AI Act affects every app that offers AI features, from chatbots and image recognition to automated candidate screening. For most apps it mainly means transparency duties and AI literacy within the team, but for certain use cases it brings far-reaching requirements for high-risk systems. The key is to classify your AI features correctly before you build them.

The EU AI Act in brief: timeline and status in 2026

The AI Act (Regulation (EU) 2024/1689) entered into force on 1 August 2024 and applies in stages. Since 2 February 2025, the prohibitions of certain AI practices and the AI literacy duty (Art. 4) apply. Since August 2025, obligations apply to providers of general-purpose AI (GPAI) models, i.e. the makers of large language models. Most of the remaining obligations, including the transparency duties, were originally scheduled from August 2026. Postponements and simplifications of the high-risk rules were debated at EU level. Check the current status of deadlines before starting a project.

This article is a practical overview for app providers and is not legal advice.

The four risk levels and where typical app features land

Risk levelExamplesConsequence for your app
ProhibitedManipulative techniques that harmfully influence behaviour; social scoring; exploiting vulnerabilities of specific groups; certain emotion recognition at work and in educationMust not be offered
High riskAI for candidate screening, creditworthiness assessment, evaluating learning outcomes, access to essential services, certain biometric applicationsExtensive duties: risk management, data quality, documentation, human oversight, conformity assessment
Limited risk (transparency)Chatbots, AI-generated images, audio and video, deepfakes, emotion recognitionLabelling and information duties
Minimal riskSpam filters, recommendations, image enhancement, autocomplete, translationNo specific duties, voluntary codes of conduct

Most AI features in business and consumer apps fall into the "minimal" or "transparency" categories. Things get critical when your app makes decisions about people or substantially prepares them, for example in HR, financial services or education.

Provider or deployer? Your role determines your duties

The AI Act mainly distinguishes between providers (whoever develops an AI system, or has it developed, and places it on the market under their own name) and deployers (whoever uses an AI system under their own authority in a professional context).

  • If you build an app with an AI feature and offer it to customers, you are usually the provider of that AI system, even if a third-party language model works underneath via API.
  • The model maker has its own GPAI obligations, but these do not relieve you of your obligations for the overall system.
  • If your company uses a ready-made AI tool internally, you are a deployer, with fewer but still real duties, especially for high-risk systems.
  • If an agency builds the app for you, you as the client offering it under your name typically remain the provider. Clarify documentation and cooperation duties in the contract.

Transparency duties: what chatbots and generative features must do

For the most common AI features in apps, the transparency duties under Art. 50 are relevant:

  1. Chatbots and assistants: users must be able to recognise that they are interacting with AI, unless this is obvious from the context. A clear note in the chat window is the simplest solution.
  2. Generated content: synthetic images, audio, video and text should be marked as AI-generated in a machine-readable way, as far as technically feasible.
  3. Deepfakes: anyone generating realistic image, audio or video content of people must disclose it.
  4. Emotion recognition and biometric categorisation: affected persons must be informed, provided the application is not prohibited anyway.

In new apps these requirements can be implemented with little effort. It gets more expensive when existing apps have to be retrofitted, for example to embed metadata in generated images.

Checklist for AI features in your app

  • Inventory all AI features: purpose, model, provider, data flows.
  • Assign each feature a risk level and document the reasoning.
  • Rule out prohibited practices, such as manipulative designs targeting vulnerable groups.
  • Build transparency notices into the interface and the content.
  • Provide human oversight where AI outputs have consequences for people.
  • Ensure quality: test sets, monitoring, handling of errors and hallucinations.
  • Review data protection in parallel: the GDPR continues to apply regardless of the AI Act.
  • Ensure and document the AI literacy of the staff involved.
  • Seek legal advice early if a high-risk case is possible.

On the reliability of AI outputs, read our article on avoiding hallucinations; on privacy, GDPR-compliant AI features. The article AI literacy under the EU AI Act explains training duties for teams.

Penalties, risks and consequences for your app project

The AI Act provides for substantial fines: up to 35 million euros or 7 % of worldwide annual turnover for prohibited practices, with lower caps for other violations. Graduated rules apply to SMEs and start-ups. For many app providers, however, business customer requirements are more relevant in practice than fines: companies and public buyers increasingly ask for documentation, risk classification and transparency measures before adopting an app with AI features.

Plan the AI Act classification as a fixed part of the concept phase, just like data protection. For typical features such as chat assistants, summaries or document recognition, clean documentation plus transparency notices is usually sufficient, and the extra effort is manageable. With us, this classification is part of the free analysis before the fixed-price offer, which you can request through our offer wizard.

Agentic coding workshop for your team? From €799 net – request a quote by e-mail

Conclusion

The EU AI Act does not ban AI in apps; it requires deliberate classification. For most apps this means transparency towards users, competent teams and traceable documentation. If you plan features in HR, finance or education, clarify the high-risk question early and keep an eye on current deadlines, as parts of the timeline have shifted.

Frequently Asked Questions

Does the EU AI Act apply to my app with a ChatGPT or Claude integration?

Yes, if you offer the app with its AI feature in the EU. You are then typically the provider of an AI system and must meet transparency duties in particular, while the model maker has its own obligations.

Do I have to label AI-generated content in my app?

Under Art. 50, synthetic content should be marked in a machine-readable way and deepfakes disclosed; chatbots must be recognisable as AI. Check the current application date of these duties; this article is not legal advice.

When is an AI feature a high-risk system?

Mainly when it is used in the areas listed in Annex III, such as recruitment, creditworthiness, education or access to essential services. Recommendations, translation or chat assistants without such decisions are usually not included.

What does AI Act compliance cost for an app?

For minimal-risk features or those with transparency duties, the effort is usually small and consists of documentation and notices in the interface. High-risk systems require considerably more effort for risk management, testing and conformity assessment.

This article was written by the editorial team of app-entwickeln-lassen.de – the app and AI development service of jocapps GmbH, based in Freiburg, Germany, and building apps since 2014. The content is for general information and does not replace individual (legal) advice.

You might also like

Legal, Privacy & SecurityGDPR-Compliant App Development: The 2026 Checklist

GDPR-compliant app development: a practical checklist from concept and SDK selection to operations, so data protection never becomes costly retrofitting.

6 min read
Legal, Privacy & SecurityBFSG and Apps: What Germany's Accessibility Act Requires

The BFSG and apps: which apps must be accessible since June 2025, which exemptions apply and how to check your app with a clear, practical checklist.

6 min read
Legal, Privacy & SecurityAvoiding App Store Rejection: Common Reasons and Fixes

Avoid app store rejection: the most common reasons at Apple and Google Play, how to respond correctly and a pre-submission checklist to prevent rejections.

6 min read
Legal, Privacy & SecurityLegal Notice, Privacy Policy and Terms in Your App: A Guide

Legal notice (Impressum), privacy policy and terms in your app: which disclosures German law requires, where they belong and how to integrate them cleanly.

6 min read
Legal, Privacy & SecurityTracking Consent in Apps: What Germany's TDDDG Requires

Tracking consent in apps under Germany's TDDDG: when you need consent, what a compliant consent dialogue looks like and how to start SDKs correctly.

6 min read
Legal, Privacy & SecurityData Processing Agreements (DPA) for App Development and Hosting

DPAs for app development and hosting: when you need a data processing agreement, what it must contain and how to review providers and sub-processors properly.

6 min read