New York Times ChineseMay 13, 2026, 14:37 UTC+8

China Sought Access to Anthropic’s Latest Technology but Was Rejected

中国曾寻求获取Anthropic最新技术但遭拒绝

Chinese think-tank representatives asked Anthropic in Singapore last month to give Beijing access to Mythos, and Anthropic refused; the company has limited the vulnerability-finding model to the U.S. government and more than 40 organizations.

Why it matters: HKR-H/K/R all pass: the NYT report gives the Singapore request, Mythos’s bug-finding use, and its US-government-plus-40 access scope. This is a same-day security and US-China AI access story.

Read the original ↗Export Markdown

Take

Mythos has crossed into quasi-arms-control territory. Anthropic is not selling a coding model; it is drawing a U.S.-aligned access perimeter. After the April launch, Mythos went only to the U.S. government and more than 40 organizations. Chinese think-tank representatives asked in Singapore last month for Beijing access, and Anthropic refused. That user list is too small to read as normal enterprise gating.

The NYT cites U.S. estimates that OpenAI ChatGPT 5.5 and Anthropic Mythos pushed the U.S.-China model gap from about six months to nine-to-twelve months. I don’t fully buy that gap as clean measurement; national-security briefings always carry deterrence theater. But vulnerability discovery changes the product category. DeepSeek adapting to Huawei chips helps the compute story. It does not solve access to a restricted cyber-capability model.

Source: New York Times Chinese · cn.nytimes.com
Published: May 13, 2026, 14:37 UTC+8 · May 13, 2026