AI at TalTech

TalTech’s vision

TalTech sees artificial intelligence as a transformative direction that affects the entire landscape of research and education. Our goal is to not merely be users of the technology, but responsible leaders and trendsetters in research, teaching, everyday work, as well as in shaping ethical and legal frameworks.

At TalTech, AI is not just a new tool, but an accelerator of institutional evolution that touches teaching, research, management, work processes, and infrastructure. To do so, we have established a strong coordinating structure at the university level, through which we support students, lecturers, researchers, and staff with training and adopted a carefully chosen set of AI tools along with guidelines, resources, and a support framework to ensure responsible use.

Rules in force at the university

At TalTech, the use of artificial intelligence is regulated by several agreements and guidelines. For teaching, the Guidelines for the Use of Artificial Intelligence in Teaching have been prepared, providing instructions on how to apply AI in a way that supports learning objectives and does not undermine the integrity of the learning process. In addition, the Tallinn University of Technology rules on the processing and protection of personal data must be followed, which set out the principles for collecting, storing, and processing personal data, along with agreements and guidelines established within individual faculties or units. The Education Department has also developed guidelines for adapting to AI tools in teaching and created a self-check framework to analyse the use of AI in learning.

At the national level, the Ministry of Education and Research has issued recommendations for the use of AI in education, emphasising transparency, the development of critical thinking, and protecting students from excessive automation. At the European Union level, the General Data Protection Regulation (GDPR) applies, setting rules for the processing of personal data in AI systems as well. GDPR requires ensuring that the use of AI does not violate privacy or create additional data-related risks. In addition, the EU Artificial Intelligence Act (AI Act) is being gradually enforced, introducing risk-based categories for AI solutions and bringing additional transparency and security requirements for systems used in the education sector and in research.

The use of AI at TalTech is based on the following principles:

  • Critical thinking and avoiding bias – AI outputs may not be factually accurate or neutral, therefore they must always be checked and evaluated with a critical eye.
  • Privacy and data protection – sensitive and personal data must not be entered into unsupported AI platforms. Only secure solutions permitted by the university or assessed under these guidelines may be used (e.g. Microsoft Copilot, in the future ChatGPT Edu).
  • Transparency and referencing – the use of AI tools in work must be disclosed honestly, for example in the preface, as a reference, or in a dedicated tools section.

Before using an AI solution, its compliance with these guidelines must be validated. The exception is centrally provided solutions offered by the university (Microsoft Copilot and, in the future, ChatGPT Edu), which have already been assessed as highly suitable. For all other solutions, suitability must be evaluated based on the criteria outlined in the guidelines, taking into account both data protection and confidentiality requirements. The final responsibility always remains with the user — including data security, protection of personal data, and compliance with the service’s terms and conditions. If necessary, IT user support should be consulted before adopting a new AI solution to ensure alignment with the university’s rules and legal framework.

Guidelines for Assessing the Suitability of AI Solutions

Purpose
This guideline helps TalTech employees assess the technical and legal compliance and purposeful use of new AI solutions, based on EU AI Act risk levels and GDPR requirements.
The guideline must be used by anyone who wishes to use or adopt an AI solution in professional activities that is not centrally provided or managed by the university.

The assessment takes place in two stages:

  • Part I – the solution is assessed based on six criteria (5 mandatory and 1 optional), using the traffic light principle.
    Part II – the results are summarized, and it is determined at what suitability level and under what conditions the solution may be used at the university.

Part I – Technical and legal suitability of the solution

  • Data storage location – Assesses where the data used and stored by the AI solution is located.
    🟢 Green – Data is located in the EU or in data centers within the EU and complies with EU data protection requirements.
    🟡 Yellow – Data is located outside the EU in countries with certain data protection safeguards, but not fully at the EU GDPR level.
    🔴 Red – Data is located in countries lacking adequate legal safeguards or with high political and security risks.
  • Use of data for model training
    🟢 Green – Terms clearly confirm that data is not used for model training or it is disabled by default.
    🟡 Yellow – Model training is enabled by default, but the user can disable it in settings.
    🔴 Red – Data is used for model training without restrictions and cannot be disabled.
  • Data ownership and usage rights
    🟢 Green – Data remains the property of the university or user, and the service provider does not use it for other purposes.
    🟡 Yellow – Data remains the property of the university or user, but the service provider has the right to use it to a limited extent.
    🔴 Red – By submitting data, the university or user loses effective ownership and control rights.
  • Ethical, pedagogical, and scientific suitability
    🟢 Green – Terms state that the solution avoids bias, discrimination, and unfair influence.
    🟡 Yellow – Ethical principles are present in the terms, but independent evaluations are lacking or limited.
    🔴 Red – No clear ethical guidelines exist, or there is evidence of bias/unfairness.
  • Service provider’s cybersecurity and data protection level
    🟢 Green – Holds international certifications (e.g. ISO/IEC 27001, SOC 2, GDPR compliance).
    🟡 Yellow – Partial compliance or certifications in process.
    🔴 Red – No certifications or inadequate description of security measures.
  • Integration possibilities and compatibility (optional)
    🟢 Green – Supports SSO (UNI-ID/Entra ID), SCIM, and API interfaces.
    🟡 Yellow – Supports some but not fully.
    🔴 Red – Does not support SSO or other critical integration options.

Part II – Purposeful use and risk level

  • 🟢 High suitability (green) – No criteria are red and at most one is yellow. Suitable for use also with personal data and internal documents.
    Use when: the solution is needed in the work process and personal data must also be processed.
    Examples: Microsoft Copilot (M365), ChatGPT Edu.
  • 🟡 Limited suitability (yellow) – All criteria are green/yellow, none are red. Recommended only for public or low-confidentiality data.
    Use when: you need a solution for processing public information or low-confidentiality data.
    Examples: Public ChatGPT, some cloud-based image generators.
  • 🔴 Unsuitable (red) – At least one criterion is red. Must not be used in everyday work.
    Use only when: testing with public data or exceptionally in a research project.
    Examples: AI services based in China or Russia.

Note: The traffic light suitability level indicates technical and legal compliance, but the EU AI Act risk level and the purpose of the solution must always be taken into account.

EU AI Act risk levels – must be considered in addition to the suitability level:

  • ❌ Unacceptable risk – prohibited activities
    Examples: covert psychological manipulation, unauthorized facial recognition, discriminatory algorithms in evaluation or recruitment.
    Must not be used.
  • ⚠️ High risk – directly affects rights and freedoms
    Examples: automated evaluation, recruitment systems, predicting academic progression, management of critical infrastructure.
    May only be used with highly suitable solutions and prior risk analysis.
  • ℹ️ Limited risk – transparency requirement
    Examples: chatbots for sharing information, image generators for creating learning materials, text tools for structuring lecture materials.
    Allowed only with highly suitable solutions, provided users are aware of the AI’s role.
  • ✅ Low risk – everyday aids
    Examples: email sorting, time management assistants, grammar checking in work tools.
    Suitable for regular use; limited suitability may also be used.

Tools

AI is becoming an increasingly important tool in universities, helping to improve teaching, research, and administrative processes. The following provides an overview of the main AI tools at TalTech.

Microsoft Copilot – An AI-based assistant
Microsoft Copilot is an AI assistant that helps draft and edit texts, summarize meetings, analyze data, and speed up everyday work. It can be used both as a chat environment and as an aid within Microsoft 365 applications.

Copilot (chat)

Free with a UNI-ID account. Copilot chat is a web-based AI assistant where you can ask questions, draft texts, search for ideas, and solve problems in natural language.

  • Web: open copilot.microsoft.com, sign in with your UNI-ID account, and start chatting.
  • Edge sidebar: open Microsoft Edge, sign in with your UNI-ID account, click the Copilot icon (top right), and type your query or select Compose.
  • Mobile: Copilot can be used in the Microsoft 365 mobile app after signing in with your UNI-ID account.

Copilot for Microsoft 365 (paid)

Additional licensed functionality in Office applications. Copilot 365 works directly in Outlook, Word, Excel, PowerPoint, and Teams, helping to create, analyze, and improve documents, presentations, spreadsheets, and communication.

Once the license is linked to your UNI-ID account, Copilot becomes available in the applications (deployment may take from a few hours up to a couple of days).

  • Ordering: support portal
  • Usage: open an Office app → click the Copilot icon → describe what you want to draft/analyze.

Data protection

When used with a UNI-ID account, enterprise-level data protection applies: your queries and responses are not used for training base models, and processing remains within the Microsoft 365 service boundaries.

Access to AI models via Azure

What is Azure AI?

Microsoft Azure AI enables the use and development of various artificial intelligence solutions — from large language models (LLMs) to image, speech, and data analysis. All models and services: https://ai.azure.com/catalog

Why use Azure AI at the university?

  • faster data analysis and research;
  • access to state-of-the-art models without building your own GPU infrastructure;
  • scalability;

How to use Azure AI services

1. Opening an Azure subscription
Contact the TalTech IT Department to open the required Azure subscription. Required information: organizational unit, financial source, and responsible requester. Request: https://taltech.atlassian.net/servicedesk/customer/portal/14/group/30/create/406

2. Creating admin accounts and configuring access rights
Azure administrator cloud accounts can be requested here: https://taltech.atlassian.net/servicedesk/customer/portal/14/group/33/create/374

3. Ordering and using services
After access has been created, it is possible to order and use AI/ML services and virtual resources (computing power, data storage, etc.) provided by Azure. Pricing information is available directly in the Azure AI catalog for each model.

Recommended areas of use
ChatGPT Edu

Note: ChatGPT Edu has not yet been adopted at TalTech. This guideline provides an overview of what the solution is and how it can be used in the future. ChatGPT Edu is planned to be introduced at the university in the near future.

ChatGPT Edu is an AI assistant based on the GPT-5 model – OpenAI’s most capable models for supporting work and study. It can be used for writing and editing texts, research, data analysis, coding, idea development, and speeding up everyday tasks. ChatGPT Edu is designed specifically for universities and operates in a secure environment where data is not used for training base models.

Where and how to use

  • Web: open chat, sign in with your UNI-ID account, and start a conversation.
  • Mobile: download the ChatGPT mobile app, sign in with your UNI-ID account, and use chats on the go.
  • In teaching and research: can be used for document analysis, processing research data, creating learning materials, and generating ideas.
  • Integrated into workflows: ChatGPT Edu integrates with multiple applications (e.g. Google Drive, Teams, GitHub, Outlook) to find answers directly from your files and materials.

Capabilities

  • Writing and editing: articles, essays, reports, lecture notes.
  • Analysis and research: data processing, summaries, creation of strategies or reports.
  • Code and technical tasks: code examples, debugging, interpreting system logs.
  • Collaboration: a reliable partner that provides fact- and source-based answers in a professional tone.
  • Teaching support: can create course materials, learning games, test questions, or personalized study guides.

Data protection

  • When logging in with a UNI-ID account, enterprise-level data protection applies:
  • your queries and responses are not used for training base models;
  • the ChatGPT Edu environment complies with ISO, SOC 2, and GDPR standards.

Other guidelines and useful materials

How to communicate with AI effectively?

Prompting means the skill of giving clear and purposeful instructions to an AI (e.g. ChatGPT, Copilot) in order to get an accurate and expected response. This skill is useful in learning, research, project management, and everyday work.

Basic rules

  • Be specific and precise.
    Bad: “Tell me about projects.”
    Good (student): “Make a 5-point summary of the main topics from the cybersecurity lecture.”
    Good (employee): “Make a 5-point overview of TalTech’s digital transformation projects 2020–2023.”
  • Assign a role or style.
    Student: “Act as an exam lecturer and create three control questions on machine learning models.”
    Employee: “Act as a document editor and correct the text to comply with ISO 27001 requirements.”
  • Explain context and background.
    Example: “Target group: first-year students, goal: study summary, format: slide presentation.”
  • Break the task into steps.
    Example: “First describe pros and cons, then make a recommendation, finally give an example from TalTech practice.”
  • Use examples and format.
    Example: “Write a summary in 100 words, then add three bullet points at the end.”
  • Refine and clarify.
    Example: “Please clarify point 3 in the TalTech context” or “Give an example from Estonian law.”
  • Ask for reasoning or thought process.
    Example: “First explain the reasoning, then give the recommendation.”
  • Set limits and scope.
    Example: “Write a 200-word essay” or “Provide exactly 5 points.”
  • Use reusable templates.
    Example for learning: Task: Explain the concept of [X] so that the first explanation is simplified, the second detailed.
    Example for work: Background: The IT department is preparing an information security policy. Task: Create a 7-point guideline for log analysis.

Summary

A good prompt = clear instruction + context + desired format.

  • For students: helps to learn faster and more deeply (e.g. exam revision, generating examples).
  • For employees: helps to prepare documents, analyses, and overviews more efficiently.

👉 Think of AI as a colleague or supervisor to whom you must give the same clear task as you would to a person.

Other links
AI pilots

This list brings together the AI solutions currently being piloted at TalTech.

  • Speech Technology Lab services. Info: Tanel Alumäe. Esileht | teksiks.ee
  • Intelligent Search (for university staff). An internal TalTech AI application that helps users quickly find meaningful answers from official university sources such as the intranet, guidelines, legal acts, and the public website. The user enters a question, the system searches for the relevant text passages, and generates an answer using the OpenAI GPT model (incl. GPT-4 Turbo). The result includes references to the original sources, and users can provide feedback on the relevance of the answer.
  • ChatBot OpenAI testing with IT faculty students. Info: Ago Luberg. Taltech Chat
  • IT-interest guide experiments (research project). A chatbot that helps students find a suitable IT bachelor’s programme before entering TalTech. Pilot 1: ithuviarajata.pragmatiqai.com For teachers, a supporting slide deck has also been created to use in class sessions. Guidelines for teachers/group implementersInfo Birgy Lorenz.
  • AI-Mentor at hackathon (autumn 2025). Info: Birgy Lorenz.
  • Eduflex model piloting in a course. Info: Margus Püüa.
  • Text-based chatbot in Moodle course. Info: Raivo Sell.
  • Text-based chatbot in Moodle/Discord course. Info: Ago Luberg.
  • Use of LENA in lectures. Info: Tarmo Koppel.
  • Development of a math tutor chatbot, currently collecting ideas. Info: Jüri Kurvits.
  • Janika Leoste integrates generative language models and TEMI V3 robot assistants into higher education teaching to guide students in the correct use of AI tools, enable more personalized and automated assessment, and reduce teacher workload.
    Info Creativity Matters ETIS
Examples from other universities

This section highlights selected universities in Europe that have well-structured AI ecosystems or a clear strategic focus on the application of artificial intelligence in education and research. Each example includes a relevant URL for further information.

Eindhoven University of Technology (TU/e)

Technical University of Munich (TUM)

Linköping University (LiU), Sweden

University of Helsinki, Finland

Aalto University, Finland

  • AI in Aalto
    A multidisciplinary approach combining engineering, design, and business. Strong focus on innovation and applied research.
    AI in Aalto | Aalto University

University of Tartu, Estonia

TU/e IE&IS (Industrial Engineering & Innovation Sciences)

  • AI in Education with an IE&IS Focus
    Additional emphasis on applying AI in the study of education and organizational behavior.
    (General link: TU/e AI in Education at IE&IS )

AI koolitused
For lecturers
Use of AI in teaching
AI koolitused
For students
Use of AI in learning
AI koolitused
For researchers
Use of AI in research
AI koolitused
For staff
Use of AI in the organization