Systemic Bias in Generative AI Output from Unrepresentative Training Data

Executive Summary

Generative AI models reproduce demographic, cultural, and linguistic biases when training data lacks diversity, producing discriminatory outputs in hiring and other decisions. Organisations deploying these tools face legal exposure and reputational harm without robust bias auditing and explainability controls.