AI in Aalto
Artificial Intelligence in Aalto
Artificial Intelligence (AI): the AI Act and AI literacy at Aalto University
The EU Artificial Intelligence Act (“AI Act”) has come into force on August 1st, 2024. How does it affect researchers, teachers, other employees, and students at Aalto University?
In a large organisation like Aalto University, there are different ways to interact with AI: you might be a user of Aalto approved AI systems like the majority of us, or you might be planning to take into use a new AI system, or you might be developing some new AI system or training an AI model. In this page we try to give you a quick overview of the AI Act and how it can affect your work at Aalto University.
What is the AI Act?
The AI Act identifies high-risk and prohibited AI systems to guide us in understanding when we need to be more careful in developing or purchasing AI technologies. Table 1 is providing an overview of various types of AI systems as they are classified in the AI Act.
| ⚪ General-Purpose AI models | 🟢 Minimal/no risk AI system | 🟡 Limited risk AI system | 🔴 High-risk AI systems | ❌ Prohibited AI practices | |
|---|---|---|---|---|---|
|
Examples of AI models or systems |
⚪ A (very large) AI model that can be used for various tasks and integrated into applications (e.g. LLMs) | 🟢 Spam filters, recommendation systems, spell checkers, translators, speech-to-text | 🟡 Chatbots, deepfakes, generative AI for purposes to inform the public on matters of public interest | 🔴 Profiling natural persons, biometric identification, determine access to education, learning evaluation, tools for recruitment (for a more detailed definition, see below AI Act relevant definitions section). AI system as a safety component of a product or AI system itself a product covered by other EU laws and evaluated by a third party (e.g. medical devices, toys, machineries…) |
❌ Deception, exploitation of vulnerabilities, social scoring, crime prediction, infer emotions in workspace or school, biometric categorization of special categories of personal data |
How does the AI Act affect Aalto University employees?
If you are a user of AI systems, there's no need to worry about the AI Act. Your focus should be on using tools that are approved by Aalto. As a community, it's crucial to share any issues you encounter—please report them to IT services or discuss them on Daily AI, Aalto Social. We will also be offering more training for staff on ‘AI literacy,’ as required by the AI Act. Be sure to explore the current AI training opportunities. If you're curious about the types of AI systems we use at Aalto, visit the 'AI in Aalto' page for more information
If you are developing an AI system or AI model or acquiring / deploying some existing AI technology you should consider if you are doing this for research purposes, or for other purposes. Continue reading below to understand how the AI Act affect your work.
Your task is related to research and you do not plan to make the AI model/system available to others.
What are the legal obligations outlined in the AI Act for research? How about ethics?
Summary: if you are developing an AI system or model for research purposes, the AI Act does not apply. The AI Act might apply if you are planning to make the research AI model/system available to others. Other legislations still apply, and research ethics is something to consider if you are going to work with high-risk AI systems or prohibited AI practices.
| ⚪ General-Purpose AI models | 🟢 Minimal/no risk AI system | 🟡 Limited risk AI system | 🔴 High-risk AI systems | ❌ Prohibited AI practices | |
|---|---|---|---|---|---|
| Legal obligations | No legal obligations. The AI Act does not apply! (Data protection and other legislations still apply of course). |
||||
| Ethics recommendations | ⚪ Ethical pre-review is optional and left to the researcher’s discretion. | 🟢 Ethical pre-review not needed | 🟡 Ethical pre-review is recommended | 🔴 Ethical pre-review is recommended | ❌ Ethical pre-review is highly recommended |
I am developing an AI system/model for research purposes, should I do anything in regard of the AI Act?
The scientific research and development are exempted from the scope of the AI Act. There are no restrictions to research and development of AI systems or AI models, including their output, that are specifically developed and put into service for the sole purpose of scientific research and development. This exclusion has been necessary as the aim of the AI Act is to foster innovation, respect the freedom of science and not to undermine research and development activity. However, any other AI system used for conducting research and development activity remains subject to the AI Act.
I am planning to develop an AI system/model for research purposes, should I do ethical review?
Research and development activity needs to always comply with ethical and professional standards for scientific research and with applicable EU law. While there might not be mandatory rules for doing ethical pre-review when developing certain high-risk AI systems, we recommend Aalto researchers to seek advice and eventually apply for the concise ethics pre-review process. If you are in doubt, get in touch with [email protected] .
I would like to open my AI model / make my AI system usable to others, do I have any legal obligations?
| ⚪ General-Purpose AI models | 🟢 Minimal/no risk AI system | 🟡 Limited risk AI system | 🔴 High-risk AI systems | ❌ Prohibited AI practices | |
|---|---|---|---|---|---|
| Legal obligations Please note that you might be a “provider” (you have created or modified an AI system and you made it accessible to others) or a “deployer” (you are installing an external AI system for others): different legal obligations apply. |
⚪ Information and transparency obligations. | 🟢 No obligations | 🟡 Information and transparency obligations. | 🔴 You need to follow various obligations as outlined by the AI Act. Get in touch with Aalto legal experts. | ❌ You cannot make these models available to others. |
| Ethics recommendations | See Table 2 | ||||
When the development of AI is no longer within the scientific research (e.g. you are collaborating with a company), and when an AI system or AI model is “placed on the market” or “put into service” as a result of such research and development activity, the AI Act will be applied to, inter alia, providers or deployers of such AI system or AI model.
-
Product-oriented research, testing or development activity
As regards product-oriented research, testing and development activity regarding AI systems or models, the AI Act does not apply prior the products are placed on the market or put into service. However, testing in real world conditions will not be covered by that exclusion.
-
AI systems released under free and open-source licenses
The AI Act does not apply to AI systems released under free and open-source licences, unless they are high-risk AI systems or an AI system that falls under prohibited AI practices or an AI system with transparency obligations. If you are unsure about which license to use or which outlet is the best for publishing your AI system/model, get in touch with [email protected].
In this section we have introduced some definitions from the AI Act. Expand the box below if you want to read more.
‘Provider’
- means a natural or legal person, public authority, agency or other body that develops an AI system or a general-purpose AI model or that has an AI system or a general-purpose AI model developed and places it on the market or puts the AI system into service under its own name or trademark, whether for payment or free of charge.
- A provider can be, for example, a researcher, a student or a person or an entity that develops an AI system or a general-purpose AI model and places that in the market.
'Deployer’
- is a natural or legal person, public authority, agency or other body using an AI system under its authority.
- A deployer can be, for example, a researcher, an employee, a student, a person or Aalto University itself that uses an AI system.
- When the AI system is used for a personal non-professional activity, this is not covered by the AI Act.
Prohibited AI practices
- The use of AI can be beneficial, but it can also be used for wrong purposes. The AI practice that does not respect human dignity, freedom, equality, democracy and the rule of law and fundamental rights, including right to non-discrimination, data protection, privacy and rights of child needs to be prohibited.
- Prohibited AI practices are, for example, AI systems that creates or expand facial recognition databases through the untargeted scraping of facial images from the internet or CCTV footage or AI systems that infer emotions of a natural person in workplace and education institutions, unless for medical or safety reasons.
High-risk AI systems
- AI systems that may pose a high risk of harm to health and safety or fundamental rights. AI systems can be high-risk, if AI system is used as a safety component of a product or AI system is itself a product that is covered by the EU harmonisation legislation listed in Annex I and the system needs to be evaluated by a third party (“a third party conformity assessment”).
- For example, these high-risk AI systems can be lifts, toys, machinery, radio equipment, pressure equipment, cableway installations, medical devices, in vitro diagnostic medical devices, automotive and aviation.
- High-risk AI systems can also be AI systems referred to in Annex III [a link] of the AI Act. For example, biometrics, e.g. AI systems intended for emotion recognition, and education and vocational training, e.g. AI systems that determine access or admission to education.
- The AI system referred to in Annex III is always high-risk, if the AI system performs profiling of natural persons.
- Exceptions for high-risk AI systems: The AI systems referred to Annex III are not high-risk, when they do not pose a significant risk of harm to the health, safety or fundamental rights of natural persons, including by not materially influencing the outcome of decision making.
Such AI system is not a high-risk if it:
- is intended to perform a narrow procedural task, e.g. classifies documents into categories or detects duplicates.
- is intended to improve the result of a previously completed human activity, e.g. AI system enhances the language in documents (professional tone, academic style).
- is intended to detect decision-making patterns or deviations from prior decision-making patterns and is not meant to replace or influence the previously completed human assessment, without proper human review, e.g. AI system that gives a certain grading pattern of a teacher.
- is intended to perform a preparatory task to an assessment relevant for the purposes of the use cases listed in Annex III, e.g. smart solutions for file handling or translation AI systems.
Limited risk AI systems
AI systems with transparency obligations
- The use of AI is permitted if the AI system meets the regulatory transparency requirements. Such AI systems are, for example, general-purpose AI systems (incl. generative AI system) and AI-generated content, chatbots and deep fakes.
Generating or manipulating images, audios or video content generating a deep fake
- If you are using an AI system to generate or manipulate image, audio or video content that constitutes a deep fake, you must disclose that the content has been artificially generated or manipulated.
- Deep fakes are AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places, entities or events and would falsely appear to be authentic or truthful to a person.
- If the generated content is a part of artistic, creative, satirical, fictional or analogous work or programme, there is only a need to disclosure that it is AI-generated or manipulated content in an appropriate manner that does not hamper the display or enjoyment of the work, including its normal exploitation and use, while maintaining the utility and quality of the work.
Generating or manipulating text for informing matter of public interest
- If you are using an AI system to generate or manipulate text that will be published to inform the public on matters of public interest, it must be disclosed that this text has been artificially generated or manipulated.
- The information needs to be presented to the natural persons in a clear and distinguishable manner at the latest upon the first interaction or exposure. The information must comply with the applicable accessibility requirements.
- If the AI-generated text has undergone a human review or editorial control, and a person or entity has editorial responsibility for the publication, there is no need to disclose that the text is AI-generated.
Minimal/ no risk AI systems
- There are no requirements for the use of AI systems that pose only minimal or no risks, e.g. video games and spam filters.
Your task is not related to research, but you are developing or deploying AI systems for use at Aalto.
If your task is not research, then you might have a few more legal obligations, as outlined in Table 3. You do not have to assess these obligations yourself, just get in touch with Aalto legal services.
I am a student, what should I know about the AI Act?
Most of the above applies also to students: for educational purposes you can explore various AI systems, just be mindful if you are working on releasing an AI system/model for others to use. Students (and teachers!) are not left alone in this journey with AI technologies, get in touch with student services if you have questions, comments, or ideas on using AI for educational purposes. Please familiarize yourself with the guidance Guidance for the use of artificial intelligence in teaching and learning at Aalto University | Aalto University
Conclusion and future work
We outline here a simple and practical summary of the AI Act and how it can affect some of us in the Aalto community. The AI regulatory landscape is still being clarified in EU and the rest of the world. Let’s remember that the main goal of the AI Act is to help us as European citizens and protect our human rights, not to create blockers or impediments to research, innovation, and education. Let’s keep the discussions ongoing and keep this page up to date with relevant materials.
References and other materials to explore
- Do you miss basic AI literacy? Take the Elements of AI course.
- Responsible use of Artificial Intelligence in the research process
- Guidance for the use of artificial intelligence in teaching and learning at Aalto University
- AI and Research Work - Useful learning materials.
- UNESCO collection on Artificial Intelligence in Education (for students, teachers, and researchers)
- The AI Act in the Official Journal of the European Union
- AI projects happening at Aalto (requires login)
Responsible use of Artificial Intelligence in the research process
How to ensure research integrity and responsible conduct of research when using AI? How should we interpret the concepts such as reproducibility; bearing responsibility on the correctness of the results presented; respecting the authorship of others, and data protection?
Guidance for the use of artificial intelligence in teaching and learning at Aalto University
Guidance for the use of artificial intelligence in teaching and learning at Aalto University.
Roosa Hasu
Legal Counsel
Kristiina Kemetter
Päälakimies
Maria Rehbinder
Senior Legal Counsel
Sirpa Syrjälä
Data Protection Officer
Laura Päkkilä
Projektityöntekijä, tietoturvallisuusjuristi
- Published:
- Updated: