# Claude resisting certain topics without tripping safeguards — retained record

> Retained reddit record from reddit:ClaudeAI.

Canonical: https://highsignal.app/data/records/e9a952b5d1a5cf1e

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:ClaudeAI)  published 2026-10-24 17:52 ingested 2026-10-25 02:36

[original source ↗](https://www.reddit.com/r/ClaudeAI/comments/1xef4h9/claude_resisting_certain_topics_without_tripping/)

## retained text



```
On my day job, Claude made entire FPGA/DSP projects work from start to finish.  
  
However working on an LLM adjacent project I noticed how there was no progress. Claude was busy, and kept offering more or less relevant ideas, but there was no solid progress on anything. I saw somewhere that Anthropic restricts the use of their models for LLM development, but my project wasn't about that, yet close enough (https://github.com/pcdeni/CaSA). The frustrating part is the lack of clear refusal. It messed up the project to the extent that took many audit rounds from the chinese models to get it back on track.

A few months ago Metatrader introduced their MCP and I started playing with it, but again there was that same resistance from Claude. Switching to the chinese models I could again see the distance between the project's intended aim and where the nearest artifacts were, but at least wasn't too far in to start over. Which is odd, because I have not come across anything about Anthropic restricting it for finance.

I am curious, did anyone experience a similar resistance on any topic?
```



document fetched 2026-10-25 00:21 · canonical https://reddit.com/r/ClaudeAI/comments/1xef4h9/claude_resisting_certain_topics_without_tripping
