O’Neil’s Weapons of Math Destruction argues that mathematical models are not inherently objective instruments, but can become systems of automated inequality when they are opaque, scalable and damaging. Her central concept, the Weapon of Math Destruction, names an algorithmic model that hides its assumptions, expands across large populations and produces harmful feedback loops. O’Neil does not reject mathematics or data science; rather, she insists that models are human artefacts shaped by choices about what counts, what is ignored and what success means. The case of Washington, D.C.’s teacher evaluation system illustrates this clearly. A teacher could be dismissed on the basis of an algorithmic score that claimed to measure “value added,” even when principals and parents recognised her as effective. The model’s authority came from its mathematical appearance, yet its internal logic was inaccessible and its errors difficult to challenge. O’Neil shows that similar systems operate in policing, hiring, credit, insurance and education, often using proxies such as postcode, credit history or behavioural data to reproduce existing class and racial disadvantage. Their danger lies not only in bad prediction, but in their capacity to shape the reality they claim merely to measure: a low score can deny work, deepen poverty and then confirm the model’s original suspicion. Her conclusion is therefore ethical and democratic: algorithms must be audited, contested and governed, because mathematical power without accountability becomes a political weapon.