awful-ai

Awful AI is a curated list of real-world cases where artificial intelligence has been used in ways that cause harm, reflect bias, or raise serious ethical concerns. The project was created to publicly track these uses and raise awareness about the risks of AI deployment in society. It is a reference document and catalog, not a piece of software.

The list is organized into categories. The discrimination section covers AI systems shown to encode racial, gender, or socioeconomic bias, such as facial recognition tools that perform poorly for people with darker skin, an image classification program that labeled Black faces incorrectly, and a UK grade-prediction algorithm that disadvantaged students from lower-income backgrounds. The disinformation and fakes category covers AI-generated deepfakes, synthetic media used for propaganda, and automated influence campaigns. The surveillance section documents cases where AI has been used to track, score, or monitor people without meaningful consent.

Other categories include social credit systems, misleading or fraudulent AI-powered products, the contribution of large AI model training to climate change, and autonomous weapon systems. The README notes that AI systems often amplify existing social biases even when trained on balanced data, and that the field is susceptible to attacks and difficult to control.

A second section of the list covers contestational AI: research and technical projects aimed at exposing, auditing, or countering harmful AI applications.

The repository accepts contributions via pull request and has been formally published on Zenodo so it can be cited in academic work.

The full README is longer than what was shown.