Chapter 5: Security
When you know the rules and guidelines for safe use of AI, you can use it with confidence.
Briefly about security
When you use AI at SDU, it is important to stay on the safe side. You therefore need to know the rules and guidelines that apply to safe use of AI - especially when it comes to which data you may use as input.
Fortunately, it is not as daunting as it may sound: you already know many of the principles from GDPR courses and other security material. But it is still worth getting clear on the most important rules of thumb, so you can use AI safely and responsibly.
To work safely with AI, you need to:
- Know the general principles for safe use of AI at SDU
- Be aware of your responsibility to use AI safely
- Know which types of data you may use as input for AI at SDU
Remember that you can always check the current rules and guidelines on the IT service page.
Data types
When you use AI technology at SDU, you need to pay attention to which types of data the different technologies are approved to handle. Not all AI tools may be used for all types of information:
- At level 0, you find publicly available information. This may, for example, be news articles, content from sdu.dk, or other material that is already publicly available online. As a general rule, this type of information may be used in all AI technologies approved for use at SDU.
- At level 1, you find internal information and ordinary personal data. This may, for example, be basic employee data, student ID numbers, exam numbers, project documents, internal SDU documents, or meeting transcripts. Here, you need to be more alert, because only some AI technologies are approved for this type of information. This is where you especially need to use your professional judgement and make sure that the tool may be used for the specific purpose.
- Confidential and sensitive personal data require particular caution. That is why they are levels 2 and 3. This may, for example, be health data, information about job applicants, minutes from closed meetings, trade union membership, or other information that requires extra protection. At SDU, this type of information is currently approved for use only in Copilot.
You can stay updated on which IT systems are approved for what here.
Copilot: The safe choice
As you can probably hear, Copilot is the safe choice at SDU - at least for now. Right now, Copilot is the only AI tool approved to handle all data types at SDU. When you use Copilot, you are therefore on the safe side.
If you use other chatbots, you need to be aware that you may not necessarily use the same material as input. Copilot may also be used for sensitive and confidential information, because the tool is part of the same closed Microsoft environment as Outlook, OneDrive and SharePoint. In short: you can confidently use Copilot in your work when you are signed in with your SDU account.
Conclusion
You can now feel confident when you use AI at SDU - especially if you use Copilot. Even if you know the rules for safe use, there will still be grey areas where the rules fall short. In those cases, you need to use your judgement and ethical sense instead. Buckle up: now we move on to the next and final chapter of this course, which is about ethics.
Rules of thumb to stay on the safe side
- Ask if you are in doubt (better once too often)
- Copilot is a safe recommendation
- If the chatbot you use is approved only for public information, what you use it for should be something you could shout across Campus Square!
Do you have any questions?
You are always welcome to contact us if you have questions or need sparring about the use of AI.
Course overview
- Introduction
- Chapter 1: What is generative AI?
- Chapter 2: Delegation: What can – and will – you use AI for?
- Chapter 2.5: Intermezzo on Personalisation
- Chapter 3: Instructions
- Chapter 4: Monitoring
- Chapter 5: Safety
- Chapter 6: Ethics
- Chapter 7: Conclusion: Continue learning about (and with) AI…
- Chapter 8: A short, engaging quiz about AI