Skip to main content

The EU AI Act for Software Teams: A Practical Primer

By the OnlineAppCreator editorial team4 min read

Short answer

The EU AI Act regulates AI systems by risk: a few practices are prohibited, high-risk systems face strict requirements, systems such as chatbots and generated content carry transparency duties, and most other software has no specific obligations. Software teams should classify every AI feature early, clarify whether they act as provider or deployer, and build documentation, logging and human oversight into the development process.

On this page
  1. The risk-based structure
  2. Provider or deployer
  3. What high-risk means for engineering
  4. Transparency duties most teams will meet
  5. AI literacy
  6. A compliance workflow for software teams
  7. Do not forget the rest of the law

The EU AI Act, Regulation (EU) 2024/1689, entered into force in August 2024 and applies in stages. For many software teams it still feels abstract, because most discussion focuses on lawyers and policy. Yet many of its requirements land directly in engineering work: logging, documentation, data governance, human oversight and transparency features in the user interface. This primer explains what software teams need to know and how to build compliance into normal development. It is an engineering overview, not legal advice; involve your legal or compliance function for binding assessments.

The risk-based structure

The Act sorts AI systems into categories with different obligations. Prohibited practices include, for example, social scoring and certain manipulative or exploitative techniques; these have applied since February 2025. High-risk systems are either safety components of regulated products or systems used in specific areas listed in Annex III, such as employment and worker management, access to education, access to essential public and private services like credit or benefits, law enforcement, migration and the administration of justice. Transparency obligations apply to systems that interact with people, such as chatbots, and to systems generating synthetic content. Everything else, which covers a large share of business software, carries no specific obligations under the Act beyond general AI literacy, although other laws such as the GDPR still apply.

Separate rules apply to providers of general-purpose AI models, the large foundation models offered by major vendors. Most software teams use these models rather than provide them, so those obligations fall mainly on the model vendors.

Provider or deployer

Your obligations depend on your role. A provider develops an AI system and places it on the market or puts it into service under its own name. A deployer uses an AI system under its authority in a professional context. A software company that builds a hiring tool and sells it is a provider; the company using it to screen applicants is a deployer. Teams building internal AI tools can be both. Roles can also shift: substantially modifying a high-risk system or changing its intended purpose can turn a deployer into a provider. Clarify roles for every AI feature, because they decide who must produce the technical documentation.

What high-risk means for engineering

If a system is high-risk, providers must implement a risk management system, apply data governance to training, validation and test data, maintain technical documentation, enable automatic logging of events, provide instructions for use to deployers, design for effective human oversight, and achieve appropriate accuracy, robustness and cybersecurity. Before placing it on the market, a conformity assessment is required. Most of these items translate directly into engineering artefacts: versioned evaluation sets, logs with retention rules, documented model and data choices, and user interfaces that let a human understand and override outputs.

The Act originally scheduled most high-risk obligations to apply from August 2026, with a later date for AI embedded in regulated products. The European Commission has since proposed adjusting some of these timelines; check the current status before planning, and do not treat a postponement as a reason to delay design decisions that are expensive to retrofit.

Transparency duties most teams will meet

Even without high-risk systems, many teams are affected by transparency rules. People must be informed when they interact with an AI system, unless it is obvious from the context. Synthetic audio, image, video or text content must be marked in a machine-readable way where the Act requires it, and deepfakes must be disclosed. For product teams this means concrete interface work: a clear notice in chat interfaces, labels on generated content and technical marking where applicable.

AI literacy

Since February 2025, providers and deployers must take measures to ensure a sufficient level of AI literacy among staff who operate or use AI systems on their behalf. For development teams, practical training on how the AI tools and features they build and use work, where they fail and how to handle data responsibly is a sensible way to meet this. Document the training.

A compliance workflow for software teams

  1. Keep an inventory of all AI features and AI tools, including their purpose, models and data.
  2. Classify each one: prohibited, high-risk, transparency obligations or minimal, and record the reasoning.
  3. Determine your role as provider, deployer or both.
  4. For high-risk features, plan documentation, logging, data governance and human oversight as part of the backlog, not as an afterthought.
  5. Implement transparency notices and content marking where required.
  6. Review the classification whenever the purpose, users or model of a feature changes.

Central infrastructure makes several of these steps easier. A governed AI platform such as KalaidoCreate provides logging, access control and compliance features across use cases, which helps with documentation and oversight. Data governance often starts with minimising personal data; local anonymisation with Textwash Pro removes identifiers before text reaches a model. For related topics, see AI governance for development teams, AI development for government and data anonymization before AI processing.

Do not forget the rest of the law

The AI Act sits on top of existing regulation. The GDPR still governs personal data, product safety and sector rules still apply, and accessibility law affects user interfaces. Fines under the AI Act are significant, reaching up to 35 million euros or 7 percent of worldwide annual turnover for prohibited practices, but the larger practical risk for most teams is having to rebuild a product because logging, oversight or documentation was not designed in. Teams that treat compliance as part of their definition of done avoid that. When OnlineAppCreator takes a product from idea to operation, the AI Act classification is part of the first scoping session for exactly this reason.

Frequently asked questions

Does the EU AI Act apply to companies outside the EU?

Yes, if they place AI systems on the EU market or if the output of their AI systems is used in the EU, similar to the reach of the GDPR.

Is a customer service chatbot high-risk under the AI Act?

Usually not. It typically falls under transparency obligations, meaning users must be told they are interacting with an AI system.

Are coding assistants regulated by the AI Act?

Using a coding assistant to write software is generally not a high-risk use. The AI literacy obligation and your internal policies still apply.

When do the high-risk obligations apply?

They were scheduled to apply from August 2026 for most Annex III systems, with later dates for AI in regulated products. Proposed adjustments to these timelines exist, so check the current legal status.

Want this done right, the first time?

OnlineAppCreator takes products from idea to operation. Tell us where you stand and get a concrete plan in a free 15-minute call.