Skip to main content

What is the EU AI Act?

The AI Glossary

In our classes, we often pause to explain the ideas behind AI and the language used to describe it. These posts grew out of those conversations. They’re for anyone who would like to understand AI better, with no technical background assumed.

The EU AI Act is the European Union’s legal framework for artificial intelligence. It establishes different obligations for different kinds of systems, uses and organizations. It is not a single approval process for all AI, and it operates alongside other laws, including data-protection and medical-device requirements.

The Act distinguishes roles. A provider develops or places a system on the market under the relevant conditions. A deployer uses it within an organization. Other roles include importers and distributors. The obligations depend on the role, the system and its intended use. Some organizations outside the EU can be within scope, but the territorial rules and exceptions need to be checked rather than reduced to “anything used anywhere in Europe.”

Certain practices are prohibited. Other uses are classified as high risk and face requirements concerning risk management, information, oversight and monitoring. There are also transparency obligations and separate rules for general-purpose AI models: models that can support a range of tasks.

Under the official implementation timeline as of November 2026, amendments have changed some earlier deadlines. Prohibitions and AI-literacy provisions began applying in February 2025, and important general-purpose-model provisions in August 2025. Most of the remaining generally applicable rules, including major transparency provisions, began applying in August 2026. Some specific transitions and additional prohibitions have a December 2026 date.

The high-risk rules for the uses listed in Annex III are scheduled to apply from 2 December 2027. The high-risk rules connected to regulated products under Annex I are scheduled to apply from 2 August 2028. Annexes are schedules attached to the law specifying covered areas and conditions. These dates should be checked again before a procurement or publication decision; older summaries may show superseded timelines.

Health care is not treated as one undifferentiated category. A medical-device function, an emergency triage application and an administrative assistant may fall under different provisions. Being used in a hospital does not by itself settle the classification.

For a health organization, begin with an inventory of functions and intended uses, identify its legal role and examine which obligations apply. A vendor’s statement that it is “AI Act ready” is not a substitute for that analysis. The Act provides a legal structure; the organization still needs evidence that its particular service is appropriate and safe. This glossary explanation is an introduction, not a legal determination for an individual product.

Sources: European Commission — AI Act Implementation Timeline; European Union — Regulation (EU) 2024/1689, consolidated text; European Commission — AI Act, Article 113

To learn more, take a look at the AI in Health Care Certificate of Specialization from the Advanced Learning Academy at Harvard T.H. Chan School of Public Health. You can take an individual course or work toward the full certificate.


AI is a fast-moving field. This explanation reflects our class discussions and related material at the time of writing. Terminology, capabilities and tools may change.

For suggested additions or corrections, please email me at tpanch@hsph.harvard.edu.

Dr. Trishan Panch
Program Director, AI in Health Care
Harvard T.H. Chan School of Public Health — Advanced Learning Academy