"...we have found strong evidence that xAI, DeepSeek, Anthropic, and
OpenAI models tend to discuss controversies from their respective companies in a differentially positive way compared to others. Meanwhile, we found no such evidence for Alibaba, Meta, and Google."
"...we have found strong evidence that xAI, DeepSeek, Anthropic, and OpenAI models tend to discuss controversies from their respective companies in a differentially positive way compared to others. Meanwhile, we found no such evidence for Alibaba, Meta, and Google."
(The magic chart is Figure 2 on page 5.)