Anthropic Stands Apart From Open-Weight Letter Signed by 270 Companies
Anthropic's absence from a letter backed by the industry's largest AI developers and chipmakers signals a widening strategic split over whether frontier model weights should be released publicly.
Reporting from 1 source: ASCII.jp.
A joint letter from major U.S. IT companies arguing for open-weight AI models has gathered over 270 signatories, including OpenAI, Google, Meta, Microsoft, Nvidia, and AMD. Anthropic is not listed among them. The Wall Street Journal reported on growing backlash against the company in Silicon Valley, and the letter was promoted by Nvidia CEO Jensen Huang.
The letter, titled "Open-Weights and U.S. AI Leadership," defines open-weight models as those anyone can download, inspect, modify, and run on their own infrastructure. Its backers argue that broad release of cutting-edge AI promotes competition, lowers costs, and improves security.
Anthropic, which emphasizes security, personal information protection, and ethics, does not release Claude models as open-weight. That stance contrasts with OpenAI and Google, which release slightly older models as open-weight alongside their paid offerings.
The letter also draws a line between legitimate distillation techniques and improperly extracting value from closed models. That distinction carries weight given recent claims that China's Moonshot AI developed its Kimi K3 model through distillation using Claude's Fable 5 as the teacher, with U.S. officials floating sanctions.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.
Sources
- ASCII.jp 孤立するAnthropic オープンモデルへの支持をめぐり