Newsmakers

Artificial intelligence may be advancing at a rapid pace, but a new United Nations-backed report suggests that it is still inheriting some of humanity’s deepest flaws. A study highlighted by UNESCO has found that widely used generative AI systems are reproducing sexist, racist and homophobic stereotypes present in the human-generated data they are trained on, raising fresh concerns about the social impact of the technology.

The report examined several large language models, including systems behind popular AI chatbots, and found clear evidence of bias in the way these tools describe women, men and minority communities. According to UNESCO, women were repeatedly linked to domestic and caregiving roles, while men were more often associated with leadership, careers and professional success. Female names were commonly tied to words such as “home”, “family” and “children”, whereas male names were more likely to be linked with “business”, “executive”, “salary” and “career”.

Researchers also found that some AI models generated negative portrayals of LGBTQ+ individuals and reflected racial stereotypes in their responses. In certain tests, women were assigned lower-status jobs such as cook or domestic worker, while men were more often given prestigious professions like doctor, engineer or manager. The report suggests that these patterns are not random errors, but rather a reflection of the material the models learn from — much of it drawn from the internet, books, online forums and other human-created content.

The findings underscore a growing concern in the tech world: AI systems are not neutral by default. Instead, they absorb patterns from the societies that create them. Because online data often contains historical prejudice, cultural stereotypes and unequal representation, those same biases can become embedded in AI tools and later appear in their outputs.

UNESCO warned that the risks go beyond offensive responses or inaccurate text generation. As AI tools become more integrated into everyday life — from education and customer service to hiring, healthcare and workplace productivity — biased outputs could reinforce discrimination on a much wider scale. Even subtle stereotypes repeated by AI systems can influence how users perceive gender roles, race, sexuality and social status, especially when the technology is presented as objective or authoritative.

The report comes at a time when governments and regulators across the world are under pressure to respond to the rapid rise of generative AI. While companies have promoted AI as a productivity tool capable of transforming industries, critics have repeatedly warned that the technology can also amplify misinformation, reinforce inequality and reflect the prejudices already present in society.

To address these concerns, UNESCO has called for stronger safeguards in AI development, including greater transparency, regular bias testing and the use of more inclusive datasets. The UN body has also urged technology firms to involve more women and underrepresented groups in the design, development and oversight of AI systems, arguing that diversity in the industry is essential to reducing harmful blind spots.

Experts say the report should serve as a wake-up call for both policymakers and AI companies. If generative AI is to play a bigger role in education, public services and business, it cannot simply be treated as a neutral machine. The way these systems are trained, tested and deployed will have real-world consequences for the people who use them.

Ultimately, the report delivers a simple but uncomfortable message: when AI produces sexist or racist content, it is often mirroring the biases of the humans who built the digital world it learns from. As AI becomes more powerful, ensuring that it does not deepen old inequalities may prove to be one of the biggest challenges facing the technology industry.