It's illegal to build a gaydar in the EU (artificialintelligenceact.eu)

The recent updates to the EU's AI Act have declared it illegal to develop AI systems that could implement "gaydar" technology—tools designed to infer sexual orientation through data analysis. The Act classifies AI systems based on risk, with unacceptable risks, like those manipulating or categorizing individuals based on sensitive attributes, outright prohibited. This sets a significant precedent within the AI and machine learning community, emphasizing the need to safeguard individual rights against invasive technological applications. For developers, the implications are multifaceted. High-risk AI systems will now face stringent requirements, such as establishing a risk management system, ensuring data governance, and enabling human oversight. General Purpose AI (GPAI) models must adhere to documentation and transparency obligations, including revealing training data sources and conducting assessments for systemic risks. These regulatory measures reflect a growing trend toward ethical AI development, forcing providers to consider the societal impact of their innovations while steering clear of technologies that could harm vulnerable populations.