Chinese AI researchers are winning the open-discourse war as OpenAI and Anthropic employees go silent on X

Chinese AI researchers are winning the open-discourse war as OpenAI and Anthropic employees go silent on X

When Silicon Valley says Chinese AI models like DeepSeek R1 "came out of nowhere," the follow-up question matters: out of whose view? On X, the people explaining how those models work are increasingly Chinese researchers. Their Western counterparts largely are not.

Over the past year, researchers at labs including Moonshot AI, Minimax, and DeepSeek have joined X to discuss their work, recruit talent, and shape how the global community understands Chinese AI capabilities 1. Moonshot AI, the company behind Kimi K3, has around 30 accounts on X from people who claim to be currently affiliated with the company, including two cofounders 1. Even several employees at DeepSeek, who reportedly cannot leave China because the government confiscated their passports, regularly post about AI developments and share job listings 1. At the same time, employees at OpenAI and Anthropic have grown more reluctant to share their thinking online as their employers have scaled into large companies, says Tiezhen Wang, an independent AI analyst and former researcher at Hugging Face, who suspects these researchers treat their model architecture and training methods as something of a secret 1.

This is not a social media trend. It is a shift in who controls the conversation about AI progress, and it carries consequences for how the world judges who is actually ahead.

The shift was not accidental. Wang personally advised Moonshot during a one-off consultation in 2025 to push more of its researchers to the front on social media 1. The result is visible: Moonshot's researchers discuss big releases, share incremental research papers, and build relationships with Western researchers, all in public 1. Western AI researchers, eager for direct information about Chinese labs, have welcomed the influx 1.

The reason these researchers chose X over domestic platforms says something about technical discourse inside China. Zhihu, once a Quora-like destination for educated professionals, has pivoted to fiction content since 2020, and expert contributors have gradually left 1. Xiaohongshu, or RedNote, reaches a young, urban, and educated audience but is less technical 1. When Moonshot published a research paper that attracted praise from Elon Musk, a current Moonshot researcher, who was not authorized to speak to media, said commenters on Chinese platforms focused on the authors' backgrounds and ages rather than the technical work, calling the discussion "very superficial and gossipy" 1.

The migration was triggered by DeepSeek R1, which went viral at the start of 2025 and put a spotlight on China's AI industry 1. Meng Fanqing, cofounder of the startup Evolvent AI and a former intern at Moonshot who worked on the Kimi K2.5 model, started actively posting on X in mid-2025 1. X itself is owned by Elon Musk, whose own AI company, xAI, competes in the same race these researchers are discussing 2.

Wang argues this trend is a net benefit for the United States because it reduces blind speculation about what Chinese labs are building 1. More information does beat less. But Wang's US-specific framing reveals the strategic calculus at work: he sees Chinese researchers speaking openly as beneficial to American interests, not as neutral transparency. The asymmetry is the actual story. Western labs that once set the terms of AI discourse through blog posts, research papers, and engineer commentary have pulled back. Chinese labs have not merely filled the gap. They have structured it, with researchers who discuss big releases, share incremental findings, and cultivate relationships with the exact Western audiences who allocate capital, choose models, and draft policy. When only one side of a competition explains its progress in public, the consensus about who is ahead bends toward their framing.

The AI race will be decided by capability. But the perception of who leads, which guides investment and policy worldwide, is being written right now on a platform where only one side is showing up to talk.