AI giants are facing the challenge of ensuring that artificial intelligence models reflect the world's diversity without being excessively politically correct. There is a concern that AI, built on potentially biased information, is at risk of automating discrimination. The quality of AI's decision-making depends on the input it receives, which can contain both useful and prejudiced content. It is dangerous because people are increasingly relying on AI software. Bias in AI can be reinforced by bias in human culture. Ensuring AI reflects human diversity is not just a political choice; it is crucial for preventing discrimination in AI applications like facial recognition.

However, there is a fine line between adequately reflecting diversity and being overly politically correct, as seen when Google's image generator produced an image of German soldiers from World War Two, including individuals of different races. Technological solutions to bias are limited, as generative AI models cannot reason about biases nor take action to address them. The responsibility to ensure appropriate and unbiased AI output currently lies with humans. Efforts to address bias in AI models are ongoing, but it is a complex task given the rapid pace at which new models are introduced. Methods like algorithmic disgorgement and fine-tuning are being explored, but their effectiveness is uncertain. Some believe that bias is inherent in both humans and AI and cannot be completely eradicated.