Anthropic 公布封禁潮数据：数十万假号被清
Anthropic 称境外实验室用数十万假账号盗取 Claude 输出训练自家模型，新防御系统已封禁数万账号、数据窃取量降九成。
ANTHROPIC JUST EXPOSED FOREIGN AI LABS STEALING CLAUDE AT INDUSTRIAL SCALE. hundreds of thousands of fake accounts, stolen payment details, false identities, resold access in blocked regions, and one network running 20,000+ accounts at once all to harvest Claude's reasoning and train their own models on it so Anthropic released a new defense system, and the results are wild: → 90–99% less data harvested → tens of thousands of accounts banned, often within 2 hours → "trillions of tokens' worth of paying traffic" turned away the daily average dropped from 8,800× to 135×, and on some days the leak was too small to even measure
这条是引用着另一条写的：x.com/anthropicai/status/2109795898909950247