Race, Racism, and Racial Inequalities in Generative Artificial Intelligence

Murthy argues that Generative AI (GenAI) is not merely suffering from biased training data but actively functions as a "racial project" that reorganizes and redistributes resources along racial lines. Drawing on Omi and Winant’s racial formation theory and Stuart Hall’s encoding/decoding model, the research posits that GenAI has become a biased interlocutor that mediates social meanings, often stripping away guardrails regarding race. To test this, Murthy conducted small-scale empirical thought experiments to audit GenAI models including ChatGPT, DeepSeek, and MetaAI. The results reveal that despite safety protocols, these models can still be manipulated to produce synthetic racist disinformation, such as fabricating news reports about immigrant violence in Europe. Furthermore, text-to-image prompts demonstrated persistent bias; for example, prompts for "British person" or "American" consistently rendered white subjects, while ethnic queries relied on stereotypes. The presentation concludes that GenAI operates as a neocolonial force, refashioning the world through Eurocentric signification, while also noting the institutional censorship faced by researchers studying these inequalities in the current political climate.