Skip to main content
To KTH's start page To KTH's start page

Ethical issues and legal matters

The use of generative AI has several ethical and legal implications that must be considered and addressed, as with any other digital tool at KTH. However, for generative AI it is more difficult than usual due to the complexity of the large language models and the usefulness of the technology.

What different issues can occur when using generative AI, and why do they need to be considered? We will address the most common ethical and legal issues linked to generative AI applications. Note that legal issues are often also ethically problematic, but the opposite is not always true. Find more practical hands-on tips on the page Use generative AI efficiently and ethically .

You can find related links under “further reading” at the end of the page.

Ethical issues

While generative AI is a powerful technology, using it raises ethical concerns that must be considered.

Concerns about created content

Some of the ethical issues when creating content with generative AI are included in the following list:

  • Distribution of inappropriate or harmful content. Generative AI lacks empathy and understanding of the content they create, which can cause harm in situations where such skills are required.
  • Amplification of existing bias. Generative AI can amplify certain biases that exist in the data it is trained on, leading to unfair or discriminatory outcomes. It can also reinforce user’s biases by presenting them as truths.
  • Truthfulness and accuracy. There is a risk of misinforming students, as generative AI can create false information in a seemingly confident way. Currently, there are no reliable tools that can screen AI-generated information for truthfulness.
  • False identities and deepfakes. Pretending to be someone else is increasingly simple due to generative AI applications, as is creating false identities. Teaching students to use generative AI will therefore also raise awareness about deepfakes, which can have positive and negative consequences. What are deepfakes? (businessinsider.com) .

Other important concerns

Even if the we solve the ethical concerns regarding the content, there are concerns with merely using generative AI. Here we list some of those concerns:

  • Uncertainty and accountability. It can be difficult to explain the basic principles and inner workings of generative AI, and nearly impossible to understand why specific content was created. This can lead to uncertainty and mistrust in content created by generative AI and make it harder to see who is responsible and accountable for the content.
  • Possible data privacy violations. As with any digital tool, data privacy and security issues must be considered before using it. Particularly when it comes to sensitive information or content created by students.
  • Environmental impact. Like all digital tools, generative AI requires power to train and use. If the owners of an AI want to keep it up to date, they have to add more data and retrain the model, which is power-intensive. It will also take up more server space and therefore requires more power to operate.

What can be done?

Some of these concerns can be addressed as simply as properly informing students about them, while others require larger investments of time and energy. Overall, the potential risks and ethical considerations should be fully addressed before implementing generative AI in your course. For more practical support regarding ethical use on generative AI, go to Use generative AI efficiently and ethically .

Note that since generative AI is a relatively new technology, ethical matters have not yet been fully explored. Future research should focus on addressing the unique challenges of generative AI. For example, better defining and measuring uncertainty in predictions, improving transparency, detecting biases, understanding the environmental impact, and putting in place effective safeguards against misuse.

Legal issues concerning generative AI

Legal issues can either concern the creation and training of machine learning models or the use of them. Sometimes it is hard to separate these two areas, since usage presumes an already trained model.  

Intellectual property and copyright

Generative AI has the potential to create unique and innovative content that is not possible with traditional methods. However, that content is based on works created by humans. The developers of the generative AI application may not have the rights to use or deliver the sources of the content on which the AI is based.

There is also a question of whether copyright law protects content created using generative AI, as it is not created by a human. If it is not protected, is AI-generated and then human-modified content protected? There should also be more awareness raised about where the line should be drawn between AI-generated and human-created content.  

GDPR

There are several issues concerning data privacy laws such as GDPR and generative AI. Here we will briefly explain the main ones.  

To follow GDPR, generative AI applications must allow their users to rectify, remove, and get access to any and all of their personal data. The main problem is that most large language models (LLMs), such as GPT-3, are trained on sets of publicly available data, often including personal data. There is no easy way to remove or rectify personal data from a trained AI model. Developers would have to re-train the model with new data, cleaned from all personal information, in order to address this issue properly. This would be both time- and resource consuming, and costly for the developers.  

Another legal issue to consider is that according to GDPR, anyone who is processing personal data needs to have a valid legal basis to do so. Having publicly available LLMs implies that the creators have already handled personal data during the training of the model. Therefore, they need a valid legal basis for processing, but it is unclear if they do or if they have already violated the law.  

Further reading

Ethical issues:

Copyright:  

How to Navigate the Legal Issues of Using Generative AI to Create Content (kofirm.com)

GDPR: