The AI notification feature built into Apple devices systematically exhibits racial and gender bias when summarising personal messages, according to researchers from AI Forensics who analysed over 10,000 outputs. The study found that the system disproportionately mentions ethnicity for non-white people and invents gender-stereotyped roles, like associating women with nursing and men with doctoring, when the original text is ambiguous. The findings also suggest Apple may fall under strict EU AI regulatory requirements due to the scale and reach of its model, which could trigger obligations for transparency and bias mitigation.
Date: February 12, 2026. Source: euobserver.com