Research Guides: AI in the Academic Setting: Generative AI Concerns

Overview of Ethics of AI

Specific Concerns

The following represent concerns that have been discussed in academic literature and the news media and are not exhaustive.

Accuracy: Read this September 2024 report from NewsGuard
Bias:
- ChatGPT was trained with text and code from the internet and by human feedback, a technique called Reinforcement Learning with Human Feedback. Therefore, if any of the information initially used to train it was biased and or feedback by the ChatGPT users shows bias, ChatGPT will "learn" that bias. This blog post explores more of where the bias comes from.
- Bias in models used in medicine
- Political bias in large language models
Copyright and Licensing
- Content produced by Artificial Intelligence can not be granted copyright protection. The US Copyright Office stipulates that copyrighted material must have a human author, and prompting an AI tool to create content does not count as human authorship. Generative Artificial Intelligence and Copyright Law by the Congressional Research Service from September 2023 explains the preliminary thoughts.
- Thomson Reuters wins first major AI copyright case in the US
- The US Copyright Office is issuing a Report in several Parts analyzing the issues, which will be published as they are completed.
  - On July 31, 2024, the Office published Part 1 of the Report, which addresses the topic of digital replicas.
  - On January 29, 2025, the Office published Part 2 of the report, which address the copyrightability of outputs created using generative AI.
  - On May 11, 2025, the office released the pre-publication version of Part 3 of the report. The current leader of the excecutive branch of the US, fired the Copyright Office head the next day.
- Wikimedia-Wikilegal/Copyright Analysis of ChatGPT
- Some publishers are licensing scholarly content for use as training data for LLMs: Licensing Agreement Tracker
- Copyright Clearance Center: AI, Copyright, and Licensing
Energy consumption/Environmental impact
- The carbon footprint of artificial intelligence
- Artificial intelligence's thirst for electricity
- "AI supercomputers consist of tens of thousands of AI chips connected with high-bandwidth networking equipment and consume up to dozens of megawatts of power—equivalent to tens of thousands of U.S. households." Sastry, G., Heim, L., Belfield, H., Anderljung, M., Brundage, M., Hazell, J., O’Keefe, C., Hadfield, G. K., Ngo, R., Pilz, K., Gor, G., Bluemke, E., Shoker, S., Egan, J., Trager, R. F., Avin, S., Weller, A., Bengio, Y., & Coyle, D. (2024). Computing Power and the Governance of Artificial Intelligence (arXiv:2402.08797). arXiv. https://doi.org/10.48550/arXiv.2402.08797
- A Bottle of Water Per Email
- AI's power requirements under exponential growth (RAND report, January 28, 2025)
- We did the math on AI's energy footprint (MIT Technology Review, May 2025)
Equity
- Toward ethical and equitable AI in higher education
- These women tried to warn us about AI
Ethics
- The ethics of ChatGPT- Exploring the ethical issues of an emerging technology
- In AI arms race, ethics may be the first casualty
- Teaching AI Ethics
- Millions of workers are training AI models for pennies (may require a subscription)
Inadvertent plagiarism by the human user (opinion piece from IEEE)
Overview of multiple concerns about the technology
- On the Danger of Stochastic Parrots: Can Language Models Be Too Big?
- Anatomy of an AI System
Peer Review Integrity
Privacy issues:
- As with any application/software, be aware of how they collect and use your data. This is addressed by OpenAI in their FAQ's with a link to their privacy policy.
- For more general information on AI and privacy see the Privacy and AI Governance Report from IAPP
- ChatGPT bug leaked users' conversation histories (March 2023)
- Don't tell anything to a chatbot you want to keep private (April 2023)
- Whoops, Samsung workers accidentally leaked trade secrets via chatGPT (April 2023)
- Will ChatGPT revolutionize surveillance? (April 2023)
Possible effects on human decision making
Reproducibility: the same prompt can generate different content so if using something like chatGPT for a methodology in research, it may not be replicable.

The short video below from the University of Maryland libraries summarizes the main concerns quite well.

Can generative AI use be detected?

Currently, there are no applications or programs that can definitively evaluate or alert you if the content you are reading or seeing was created by an AI like ChatGPT. There are a few out there that tell you the probability that the content was ChapGPT created such as the app GPTZero which is free and created by a computer science student at Princeton. Sensity AI has come up with an AI-powered deepfake detector. ScamAI can evaluate videos, images, voice, and text.

"As the technology stands today, it's safest to avoid automated AI detection tools completely. As of now, AI writing is undetectable and likely to remain so," frequent AI analyst and Wharton professor Ethan Mollick told Ars in July. 'AI detectors have high false positive rates, and they should not be used as a result.'" Edwards, B. (2023, September 8). OpenAI confirms that AI writing detectors don’t work. Ars Technica. https://arstechnica.com/information-technology/2023/09/openai-admits-that-ai-writing-detectors-dont-work/

As the large language models and image generators "learn" more by interacting with human users, it will become increasingly difficult for humans to detect AI content. This company, Undetectable AI, helps you humanize something you've created using AI. These articles give some more information on this topic:

There has also been some exploration into "watermarking" AI output (images, video, text). Read more about it in the articles below.