claudePOST
Anthropic 点名智谱蒸馏 Claude,数万账号被封
Anthropic 公开指控中国 AI 实验室 Z.ai(智谱)在训练 GLM-5.3 时蒸馏 Claude 推理,新防护措施让可提取的推理数据减少 90–99%,数万账号被封。
Sahil Verma
@sahilvermaai18 天前收录原推发布于 2026.10.13
Sahil Verma @sahilvermaai原推发布于 2026.10.13
Anthropic just named a Chinese AI lab in public for trying to copy Claude. It says http://Z.ai (Zhipu) tried to distill Claude's reasoning while building GLM-5.3. Anthropic's new safeguards cut what distillers harvest by 90-99%. The reasoning tokens they can pull out are now 10-100x lower, and tens of thousands of accounts got banned. Per Reuters.
x.com/sahilvermaai/status/2109869238869541368采集于 2026.10.14 10:20