Nice research and structuring into 4-tier layer. For providers like Anthropic and OpenAI, subscription is the entry point for all these, right? Besides the measures proposed in the article, can token usage % determine these clusters of accounts?
There are probably various metrics like language used to prompt the model, number of hours per day spent prompting, and many others.
There are quite a few other mitigations that could be done by providers that aren't mentioned in the article.
There are probably various metrics like language used to prompt the model, number of hours per day spent prompting, and many others.
There are quite a few other mitigations that could be done by providers that aren't mentioned in the article.