AI Models Show Divergent Web Citations, Moral Reasoning, and Scientific Ambitions

Recent developments in artificial intelligence highlight significant differences in how AI models access and process information, as well as their evolving capabilities across various domains. A study by Fudan University revealed that Western and Chinese AI search engines cite largely separate webs, with models like Qwen drawing 96% of their sources from Chinese websites even when queried in English. This divergence extends to the platforms they utilize, with Western engines favoring Reddit and YouTube, while Chinese counterparts rarely cite these, opting instead for technical sites.

Beyond information retrieval, AI's application is expanding into complex areas like scientific research and ethical reasoning. Colossal Biosciences has launched a new AI platform aimed at accelerating de-extinction efforts, signaling AI's growing role in biological sciences. Concurrently, the MoralityBench.ai benchmark is testing large language models (LLMs) against human moral psychology instruments, revealing varied responses and refusal rates among models like Muse, GPT-6 Astra, and Gemini 4 Argon. Anthropic also reported substantial revenue growth, indicating strong commercial interest in AI development.