Old Discrimination, New Speed: The UN's Warning on AI Bias
UN Deputy Secretary-General Amina J. Mohammed warns that AI, if left unguided, will not create a new utopia but will automate and accelerate our oldest prejudices.
Artificial intelligence, if left unguided, will not create a new utopia but will instead automate and accelerate humanity's oldest prejudices. UN Deputy Secretary-General Amina J. Mohammed warns that correcting this trajectory requires deliberate, conscious intervention, not just technological optimism. We are teaching our machines from the 'full archive of human behaviour' before we have agreed on the lessons they should learn.
That archive is not a neutral library of facts. It is the sum of our history, encoded in the language we use and the data we generate. The datasets that train AI models contain centuries of biased texts, skewed hiring decisions, unequal loan approvals, and prejudiced legal outcomes. These are not bugs to be patched; they are foundational features of the material we are feeding the machine. The AI is not inventing new forms of discrimination, but it is learning, perfecting, and scaling our existing ones at a speed we have never witnessed.
In her speech, Amina J. Mohammed casts this challenge in starkly human terms. When an algorithm gives old discrimination “new speed and authority,” the consequences are not abstract. This means real people being denied jobs by automated resume scanners that have learned to favor one gender over another. It means families being denied housing by lending models that perpetuate historical redlining. In **the Deputy Secretary-General's full remarks** at the AI and Human Development Regional Convening, she framed this not as a technical problem for engineers but as a human rights issue for everyone.
The essential tradeoff being made in the race for AI dominance is one of speed versus justice. The pressure to innovate and capture market share incentivizes a pace that leaves little room for the slow, difficult work of auditing datasets and building ethical safeguards. The result is that models are deployed with known flaws, under the assumption that the benefits of scale outweigh the harms of inherited bias. This calculation is a moral failure, prioritizing rapid technological progress over the dignity of those who will be marginalized by its application.
Who gets to decide our automated future?
Amina J. Mohammed points to a profound generational injustice: the people whose lives will be most deeply shaped by artificial intelligence currently have the least say in its development. The rules governing our algorithmic future are being written now, in corporate boardrooms and government committees, by a generation that will not experience its longest-term consequences. This leaves young people in the position of inheriting a world that was beta-tested on their futures without their consent.
The ghost in the machine is us. The biases AI learns are a mirror reflecting our own societal failures. Correcting the technology, therefore, requires more than a technical fix; it demands a moral and political reckoning. It means making the conscious choice to embed fairness, accountability, and human dignity into the code that will shape our lives, even if it makes the process slower and more difficult. Without that deliberate correction, we are only building a faster road to an old destination.