Generative AI is trained using vast amounts of data which often includes text data from online websites, books, articles, and social media posts. These models learn to predict the most likely sequence of words, often mirroring the prejudices prevalent online. Human bias can also be introduced during the training process, leading to AI outputs that contain sexist, racist, or other offensive content.
Artificial Intelligence (Generative) Resources by Georgetown University Library, used under CC BY-NC 4.0 / Library-specific information adapted for University of Texas at Tyler. 
Generative AI tools can support users in various aspects of the research process. However, these tools can be unreliable as they often generate false information, or "hallucinations," presented confidently. These hallucinations can include fabricated citations or facts.
AI tools have even been used to create false images or audiovisual recordings to spread misinformation and mislead audiences. Referred to as "deepfakes," these materials can be utilized to subvert democratic processes and are thus particularly dangerous.
AI generated content sometimes lacks currency as some systems do not have access to recent information. Rather, they may be trained on past datasets which generate dated representations of current events and the related information landscape.
How to Spot Deepfake Videos: 4 Simple Tips
The Top 5 Ways to Spot 'Deepfake' Videos and Images
Creating a "Lie Detector" for Deepfakes
When AI Can Fake Reality, Who Can You Trust? (Sam Gregory, TED Talk)
The Future of Fakery: Deepfakes, Generative AI & The Fight for Authenticity
Generative AI tools pose significant privacy risks because they collect and process a lot of data about users. This data can be misused, and/or sold by companies without your consent. Even if you don't directly share personal information, the patterns in your data can still reveal sensitive details about you.
Key Tips