Anthropic's February report and June follow-up detail massive campaigns by labs including DeepSeek, Moonshot, MiniMax, and later Alibaba-linked operators that created tens of thousands of fake accounts to query Claude at industrial scale. The company ties these disclosures to the need for export controls and industry-wide coordination, yet @scaling01 points out that confirming any specific outputs were actually used to train rival models remains technically unfeasible.
@scaling01 Then what does Anthropic even publish the numbers for?
@teortaxesTex That's not at all my argument.
I'm saying the numbers we see are not correct and it's much more for all of them, and that all of them are buying Claude data.
Anthropic's methods flag coordinated behavior, unusual volumes, and infrastructure signals rather than tracing individual tokens after the fact. This leaves open whether the published exchange counts fully capture distillation activity or mainly serve to highlight the regulatory stakes.
By connecting the attacks to compute requirements that current chip restrictions aim to limit, Anthropic positions the incidents as evidence that tighter rules are both necessary and currently insufficient without better enforcement tools.
Negative users accuse Anthropic of using questionable published numbers as a pretext to push AI restrictions, since the actual figures do not matter to them.
No Digg Deeper questions have been answered for this story yet.
@teortaxesTex Anthropic just want to push for restrictions, the actual numbers who distills how much doesn't matter to them
it's just very hard to prove which tokens are used for "distillation"
@scaling01 Then what does Anthropic even publish the numbers for?