Global AI trends are fundamentally changing the way we interact in areas such as knowledge, systems and decision-making. These trends include, among others:

• Multimodal models that can combine and process different data types such as text, image and sound.
• Small, highly specialised language models with a smaller data footprint.
• Edge AI, which enables locally executed AI processes directly on devices, improving speed and privacy.
• Autonomous agents, intelligent digital devices that can autonomously perform actions. The transition from traditional app-centric systems to more flexible agent-based interaction.
• A hybrid approach that unites the capabilities of AI with human intelligence to augment our cognitive abilities.
• Convergence of technologies, which occurs when AI is integrated with other advanced technologies such as robotics, biotechnology and IoT, creating entirely new applications and solutions. AI can also help convergence across existing systems, technologies and software.

There is no one tool that covers all university needs, but many different types of AI that can be curated to be used for the right types of tasks and processes.

Students and external stakeholders increasingly expect universities to:

• Offer AI competences and modern learning environments
• Work digitally and innovatively - both in operations and professionalism
• Actively participate in the development of AI solutions for society.

AI thus becomes not only an internal driver of change, but also a relevant factor in relation to the outside world and SDU's reputation.

AI is already part of everyday life for many students and employees - but in different ways and with varying degrees of responsibility. That's why it's being experienced:

• Shadow IT and unregulated use of unauthorised AI services
• Different practices for declaration and use in teaching and assignments
• Need for common standards, tools and learning spaces.

This calls for both a common culture and a common code for the use of AI. We must be open to the behaviour and offer guidance, clear governance and structure, rather than trying to control the behaviour with bans and budding, incoherent communication.

The use of AI in universities raises questions about:

• Academic integrity and cheating in education
• Bias and discrimination in AI-driven decisions
• Privacy and data protection.

An AI action plan should therefore not only be a technology plan, but also an ethical and value compass that defines SDU's position and responsibilities.

The EU AI Act, which comes into force in 2025-2026, imposes new obligations on universities:

• Risk assessment and registration of AI systems
• Transparency requirements for the use and documentation of models
• Ethical accountability in the use of (generative) AI
• Competency requirements for employees and decision-makers
• Prohibition of certain applications.

We must ensure that we have a clear and transparent governance structure that handles the legal requirements we as a university must comply with.

Like many other universities, SDU is experiencing:

• Limited technical and analytical capacity in parts of the organisation
• Lack of time and supporting structures for experimenting with new technologies
• Fragmented knowledge and practices around the use of AI.

An action plan is therefore crucial for developing our shared organisational capabilities - technical, pedagogical, legal and managerial.

Both students and researchers can AI-generate content without sufficient judgement, and AI supports mass production, but not necessarily understanding and relevance.

• If students, researchers and lecturers do not perform quality assurance to identify and manage AI-generated content that lacks academic depth, correctness or originality, it will result in a loss of academic integrity.
• If SDU fails to set the bar for what constitutes good quality in education and research, there will be a risk of quality deterioration and loss of reputation.

Mitigation can, for example, be achieved through: Courage to involve and make demands on all stakeholders in education and research to define quality concepts and standards for SDU's services. There is a need for ongoing skills development among lecturers and researchers so that they can navigate the use of AI critically and constructively.

Uncritical use of AI tools may put SDU in conflict with applicable legislation and data protection rules.

• Using external AI services without a data processor agreement, unknown training data or automated data collection can lead to breaches of GDPR and other compliance requirements.
• Lack of technical understanding can lead to errors in risk assessment or incorrect data use.

Mitigation can be done through: Differentiated security model, ongoing risk assessments and close collaboration between faculties, SDU RIO and the IT security organisation.

Lack of overview and control over AI usage in your organisation can lead to parallel systems and unauthorised use.

• Local AI initiatives risk developing in silos without coordination and common standards.
• When employees or students use AI tools without authorisation, the risk of data vulnerability and inefficiency increases.

Mitigation can be achieved through: Common guidelines, proactive communication about approved software solutions, strong communication about compliance rules, and cross-functional coordination through e.g. Communities of Practice and existing cross-functional functions.

Even great AI initiatives can fail if the organisation isn't ready to absorb the change.

• If employees don't have the time or support to test and adopt AI, the potential will remain unrealised.
• If the digital maturity of employees and managers isn't high enough for them to embrace the technology.
• If the changes brought about by the transformation are not framed in a constructive way.

Remedies can include: Management prioritisation, strategic skills development and dedicated support across subject areas and functions.

AI can lead to ethical dilemmas and legal breaches if the technology is used without careful consideration or appropriate regulation.

• The use of AI can lead to discrimination, bias and opacity, especially in decision support systems.
• Unclear rules for AI in teaching and exams can lead to distrust and cheating.
• Improper use of AI in research can compromise academic integrity.

Remedies include: A clear code of ethics, AI governance, and attention to fairness, transparency and accountability - all of which are also required of us through the European AI Act.

The implementation of AI not only involves technological and organisational changes - it also affects how the University of Southern Denmark is perceived by the outside world.

• Miscommunication or unethical use of AI can damage SDU's credibility internally and externally.
• Insufficient protection of data - especially personal data - can lead to serious breaches of trust and legal consequences.
• If AI is perceived as opaque, exclusionary or as a substitute for human relationships, it can undermine trust and SDU's value foundation.

Mitigation can be done through, for example: Open and proactive communication, clear ethics and involvement of students, employees and the outside world in the development of AI practices.

AI is often perceived as a plug-and-play solution, but without the right support and anchoring, it doesn't create value.

• The solution may lie in other digital tools, automation or better utilisation of existing solutions.
• Success with AI requires the right data, organisation, change management and skills development.
• Expecting quick wins without data preparation, governance and user engagement can lead to wasted resources and weakened trust in the technology.

Remedies include: Clear objectives and realistic implementation plans with organisational and managerial support and cross-functional strategic management of digitalisation initiatives.