The most common AI models have only been trained using online content written in English. This is because a large proportion of websites are written only in English. However, it means that these AI models are only available in English, excluding those who don't speak the language or being harder to use for those who have English as a secondary language.
As a result of this, content from AI sources is likely to be at risk of a global north and English-centric bias. This is problematic because it perpetuates the perception that research from the global south is not as significant.
It is important to always consider any potential bias in the research material that you use for your work, but if you choose to use AI at any point in your research journey, it's important to bear in mind that the content it generates will be biased from the start, due to the material that has been used to train the program.
You can read about the Library Unlimited Project here to learn more about the Library's continued efforts to decolonise our collections.