Tech Giants Race Ahead While Struggling With Fair AI

Big tech companies are in a hurry to build the next big thing, but sometimes speed leads to problems they didn’t plan for. Microsoft found this out the hard way when their AI started showing blind people with overly dramatic blindfolds instead of realistic representations. The issue? Most AI training data comes from the real world, and the real world isn’t always fair or accurate. To fix this, Microsoft turned to a nonprofit called Be My Eyes, which connects blind and low-vision users with volunteers and AI tools. They collected over 20 million minutes of video showing real blind people using canes, guide dogs, and everyday objects. After blurring faces to protect privacy, they used the footage to train their AI better. But even this approach has limits. Experts like Annie Brown, who works on reducing bias in AI, say just adding more data isn’t enough. The way images are labeled in training sets can still create unfair outcomes.Microsoft isn’t alone in this struggle. Google and others take different paths, like forming safety councils or setting core AI principles. But Microsoft’s approach stands out because it builds accountability into its AI systems from the start. Since 2002, when Bill Gates pushed for trustworthy computing, the company has shifted from rushing new features to focusing on reliability and fairness. AI’s impact isn’t just about fairness—it’s also reshaping jobs. While some companies cut roles to replace workers with AI, Microsoft’s own layoffs in 2025 were more about shifting priorities than pure replacement. Still, AI can help marginalized workers too. Microsoft’s Copilot tool, tested first by employees with disabilities, improved accessibility by offering real-time captioning, meeting notes, and sign language recognition. For neurodiverse workers, it reduced mental strain so much that some refused to give up their licenses. Inclusion matters, but it’s not just about charity. Diego Mariscal, who runs a startup accelerator for entrepreneurs with disabilities, says disabled people should be at the decision-making table—not just as afterthoughts, but as key voices shaping AI’s future. The best tech comes from diverse perspectives, not just faster code.