Artificial intelligence faces a challenge: how can optimized decision-making systems avoid perpetuating the human biases embedded in their training data? AI-enabled discrimination underscores the need to interrogate how data absorbs and reproduces societal inequities. Our research, grounded in the history of science and science studies, demonstrates that bias is not merely the product of individual beliefs or intentions but is materially distributed across databases, archives, software packages, and recording standards. Focusing on race and biometric reference databases in biological anthropology, we trace how statistical tools like the Mahalanobis distance, foundational to contemporary data analysis, are deeply entwined with the legacies of race science and colonialism. Blaming historical figures like Karl Pearson or Prasanta Mahalanobis is insufficient. We advocate for an interdisciplinary approach that critically examines the histories of datasets, technologies, and methods to disrupt the reproduction of racist harms. Our project addresses a pressing question: to what extent does the use of statistical measures perpetuate the racist ideologies they were built on, and can acknowledging their origins or modifying their use mitigate these harms? Our one-year project will refine interdisciplinary methods and expand a transatlantic network of researchers committed to confronting these challenges, advancing ethical debates on AI and data.