Bias
OpenAI, the creators of ChatGPT, explain on their own site the risk of using artificial intelligence in terms of bias:
Is ChatGPT biased? | OpenAI Help Center
Does ChatGPT tell the truth? | OpenAI Help Center
Essentially, AI like ChatGPT are prone to confirmation bias, hallucinations, and Western viewpoints, often agreeing with a user’s own perspective.
Non-GenAI users who are non-native English writers
The potential for bias does not only concern sway on the topics it generates, but also certain demographics who don’t even use it. Non-native English speakers who did not use AI were flagged with an average false positive rate of 61.3%, compared to only a 5.1% false detection rate for native English speakers in a study on AI detectors across seven widely used sites.
One of the most popular grammar websites, Grammarly, corroborates this sentiment in their article Are AI Detectors Accurate? Understanding Their Limitations as well.
When examining the indicators of human written content, detectors are often looking for lack of detail, overgeneralizations, and low sentence variability. This is called a lack of perplexity. When writing in a less familiar language, this kind of writing style can become common.