Claude Code's source code has been leaked via a map file in their NPM registry

Post by gruez » Tue, Mar 31, 2026, 3:32 PM UTC

petcat wrote: post #395

arcfour wrote: post #368

Earlier quoted context omitted.

You're perfectly free to scrape the web yourself and train your own model. You're not free to let Anthropic do that work for you, because they don't want you to, because it cost them a lot of time and money and secret sauce presumably filtering it for quality and other stuff. Stole? Courts have ruled it's transformative, and it very obviously is. AI doomerism is exhausting, and I don't even use AI that much, it's jus…

> Stole? Courts have ruled it's transformative, and it very obviously is. The courts have ruled that AI outputs are not copyrightable. The courts have also ruled that scraping by itself is not illegal, only maybe against a Terms of Service. Therefore, Anthropic, OpenAI, Google, etc. have no legal claim to any proprietary protections of their model outputs. So we have two things that are true: 1) Anthropic (certainly)…

Post by cedws » Tue, Mar 31, 2026, 3:33 PM UTC

mmaunder wrote: post #466

cedws wrote: post #69

ANTI_DISTILLATION_CC This is Anthropic's anti-distillation defence baked into Claude Code. When enabled, it injects anti_distillation: ['fake_tools'] into every API request, which causes the server to silently slip decoy tool definitions into the model's system prompt. The goal: if someone is scraping Claude Code's API traffic to train a competing model, the poisoned training data makes that distillation attempt less…

Haven’t looked at the code, but is the server providing the client with a system prompt that it can use, which would contain fake tool definitions when this is enabled? What enables it? And why is the client still functional when it’s giving the server back a system prompt with fake tool definitions? Is the LLM trained to ignore those definitions? Wonder if they’re also poisoning Sonnet or Opus directly generating si…

Post by jollymonATX » Tue, Mar 31, 2026, 3:35 PM UTC

mcv wrote: post #349

bkryza wrote: post #84

They have an interesting regex for detecting negative sentiment in users prompt which is then logged (explicit content): https://github.com/chatgptprojects/claude-code/blob/642c7f94... I guess these words are to be avoided...

I'm clearly way too polite to Claude. Also: // Match "continue" only if it's the entire prompt if (lowerInput === 'continue') { return true } When it runs into an error, I sometimes tell it "Continue", but sometimes I give it some extra information. Or I put a period behind it. That clearly doesn't give the same behaviour.

Post by nojs » Tue, Mar 31, 2026, 3:37 PM UTC

BoppreH wrote: post #103

bkryza wrote: post #84

They have an interesting regex for detecting negative sentiment in users prompt which is then logged (explicit content): https://github.com/chatgptprojects/claude-code/blob/642c7f94... I guess these words are to be avoided...

An LLM company using regexes for sentiment analysis? That's like a truck company using horses to transport parts. Weird choice.

Post by dpark » Tue, Mar 31, 2026, 3:38 PM UTC

petcat wrote: post #302

nialse wrote: post #290

Earlier quoted context omitted.

Paranoia. And also ironic considering their base LLM is a distillation of the web and books etc etc.

They stole everything and now they want to close the gates behind them. "I got the loot, Steve!" I feel like the distillation stuff will end up in court if they try to sue an American company about it. We'll see what a judge says.