Technology10/10/2025•The Verge
OpenAI is trying to clamp down on ‘bias’ in ChatGPT
OpenAI has conducted an internal "stress-test" to evaluate the political bias in its ChatGPT model. The test assessed the chatbot's responses to 100 topics, ranging from liberal to conservative viewpoints. OpenAI found that the latest GPT-5 models perform better than their older counterparts, with 30% lower bias scores. The company has taken steps to address bias, such as allowing users to adjust the tone of ChatGPT and publishing a model specification. However, the Trump administration is pressuring OpenAI and other AI companies to make their models more conservative-friendly, which could impact their efforts to maintain political neutrality.
Source: For the complete article, please visit the original source link below.
