5 ms·
https://www.anthropic.com/news/detecting-and-preventing-distillation-attacks https://www.anthropic.com/news/detecting-and-preventing-dist... Moonshot AI Scale:
by milkshakes 3mo ago
https://www.anthropic.com/news/detecting-and-preventing-distillation-attacks https://www.anthropic.com/news/detecting-and-preventing-dist...
Moonshot AI
Scale: Over 3.4 million exchanges
The operation targeted:
Agentic reasoning and tool use
Coding and data analysis
Computer-use agent development
Computer vision
Moonshot (Kimi models) employed hundreds of fraudulent accounts spanning multiple access pathways. Varied account types made the campaign harder to detect as a coordinated operation. We attributed the campaign through request metadata, which matched the public profiles of senior Moonshot staff. In a later phase, Moonshot used a more targeted approach, attempting to extract and reconstruct Claude’s reasoning traces.
- idiotsecant 3mo ago>request metadata, which matched the public profiles of senior Moonshot staff Translation: we have the machinery in place to identify our users, and actively do so.
- InsideOutSanta 3mo agoI'm assuming you posted that as evidence for the claim that "empirically, it appears that distillation of a more advanced model is a required first step", but I don't think it is. It's just evidence that Moonshot distills Anthropic's models, which, yes, they do.
- milkshakes 3mo agoit is not a required first step for training a model, sure. but that's not what i claimed. what i claimed is that is how they are so significantly _reducing the cost_ of training one! how else do you think they are doing it?