Recent criticism has targeted three AI-powered search engines of tech giants Google and Microsoft, as well as Perplexity, promoting inarguably wrong and grossly discredited race science.
Their messaging, unfortunately, amplified a dark tale: the supposed superiority of white genetic stock, according to alarm raised among experts in the field.
The Rise of Race Science
WIRED uncovered that Patrik Hermansso, a researcher with Hope Not Hate—an anti-racism organization based in the UK was pursuing his inquiry, testing an obscure dataset claiming IQ scores prove support for white supremacy.
Hermansson settled on testing the Human Diversity Foundation, which he found was funded by Andrew Conru, a billionaire tech mogul from the US, who cofounded Adult Friend Finder.
Founded in 2022, this group is treated as a continuation of the Pioneer Fund, established by Nazi apologists in 1937 to promote "race betterment."
Chilling Web Searches
During his research, Hermansson Googled IQ scores by country. He was shocked to find the feature his computer defaulted to enabling Google's Overview AI tool listed supposedly factual IQ scores for countries like Pakistan, Sierra Leone, and Kenya.
For instance, he was given shockingly high and specific IQ scores for Pakistan—information directly lifted from Richard Lynn's discredited work. Lynn, a former professor at the University of Ulster and president of the Pioneer Fund, defined the race science movement until his death in 2023.
AI-Endorsed Flawed Research
The implications of AI search engines propagating Lynn's controversial research are very troubling.
Experts say that it may have the effect of making people radicalize. Lynn's work had been exploited and used by many white supremacists and far-right hate extremists for so long as a justification for their ideologies.
Recently, Lynn's studies were cited by the Buffalo mass shooter of 2022, bringing forward the real-world danger of these narratives.
AI Tools Have Downsides
Google's AI Overviews, launched a couple of months ago to answer user queries, drew much flak. When Hermansson asked about national IQs, the AI did not draw out the questionable sources for its answers.
After complaints about this issue, Google acknowledged its shortfalls and promised that the tool would improve.
"We have guardrails and policies in place to protect against low-quality responses, and when we find overviews that don't align with our policies, we quickly take action against them," Ned Adriance, a Google spokesperson told WIRED.
Other AI Platforms Shoulder the Blame
Based on WIRED's findings, Microsoft's Copilot and Perplexity also repeated Lynn's wrong numbers when asked for IQ scores. Bing even repeated false claims by a spurious source concerning the average IQ of Pakistan, while Perplexity quoted Lynn's work outright. These results suggest that the spread of discredited race science is not confined to a given platform.
The Need for Better Regulation
According to Rutherford and Rebecca Sear, Brunel University London's director of the Center for Culture and Evolution—experts in their field, better oversight and accountability need to occur in the scientific community.
The reason for this is that Lynn's flawed research has been trotted out for so long by the academic world without proper scrutiny. This ugly narrative has thus penetrated not only the academic discourse but also AI systems.
Rutherford believes the problem isn't with AI—it lies within the academia. He continues by warning that unless asked the right questions, human and artificial intelligence alike will let false information "dribble" out.
