# Chances of Anthropic gaining access and using my AI research code? — retained record

> Retained reddit record from reddit:ClaudeCode.

Canonical: https://highsignal.app/data/records/eb567105e6dc43ef

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:ClaudeCode)  published 2026-09-20 12:45 ingested 2026-09-21 02:38

[original source ↗](https://www.reddit.com/r/ClaudeCode/comments/1wlg64j/chances_of_anthropic_gaining_access_and_using_my/)

## retained text



```
Please help me calibrate my paranoia level, because the OpenaI/math scandal made me even more paranoid than I was before. 

I'm a Machine Learning engineer working on an academic research project.

Naturally, this counts as AI research, so Anthropic can potentially flag my Claude conversation for moderation by their team if they consider it advanced enough to be dangerous.

They also have the right (according to their ToS, iiuc) to consider it as directly competing with them, which is another way they could gain access to it through moderation. 

Another scenario I'm worried about is less explicitly malicious, and is more similar to the OpenAI math scenario (as I understood it): my code ending up in Claude's training data, even though I did disable the improve-for-everyone toggle.

Obviously, working with their agent is an incredible force multiplier for research. But I am trying to figure out what are the chances that I'm better off without it, if my IP risks getting stolen (either maliciously or by ending up in the training data.)

My question is what are the chances that my intellectual property ends up being seen by their research scientists or by their agent, each could use it in different ways to scoop me.

(I am asking specifically about Anthropic because I'm using their paid tier, but the question is general and applies to all frontier proprietary agents.)
```



document fetched 2026-09-21 00:35 · canonical https://reddit.com/r/ClaudeCode/comments/1wlg64j/chances_of_anthropic_gaining_access_and_using_my
