On the IP side, About 6 months ago I used Claude to help me draft a patent application for something which I believed was novel, I verified in the USPTO as well. Then out of being busy I dropped that idea. Couple weeks ago I picked up the work again, verified with Claude for the problem statement to see if there is any existing solution. It suggested my own solution back to me which was surprising since I only use in browser and delete the history explicitly. I asked it for references or resources and it could not find any. That helped me to understand the boundaries bit better.

"When you see the “How is Claude doing this session?” prompt in Claude Code, responding to this survey, including selecting “Dismiss”, records only your rating. We do not collect or store any conversation transcripts, inputs, outputs, or other session data as part of the rating prompt itself. Unlike thumbs up/down feedback or /feedback reports, this session quality survey is a simple product satisfaction metric."

> The most common feedback on these posts is always "what did you expect?"

Is completely spot on. It's highly entertaining to me that some people can't wrap their head around that.

Use it for training/review (RLHF, whatever) and "we load the whole thread, not the specific response or surrounding few prompts, and store it for 5 years" seem completely different things. It's not obvious why being tech savvy would automatically make one aware that the 2nd version is happening.

And to be fair to them, my motivation to send quite a lot of personal data to an LLM is shockingly high.

I send medical data to openai while I never ever send any AI provider an Image of me or my wife despite me wanting to do Star Wars montages and Studio Ghibli avatars.

I disabled Gemini Activity history because they told me that they will let humans look over it, but I gave claude access to my laptop.

There is a reason why we have consumer protection. Life doesn't need to be hard by default :D

It seems like your counter-point is “they already use everyone’s data, why didn’t you just assume they would”, which is a bit weird from a victim-blaming perspective but valid nonetheless.

https://privacy.claude.com/en/articles/10023580-is-my-data-u...

The link in the post is for the API and commercial offerings. With that said, as others have already pointed out, how else could they actually make use of the feedback without the data that the feedback was given on.

They will train on it, and now competitors or future competitors will know about your niche, the specific problems you face within it, and how to solve them without doing the work that you did to get there.

That's a pretty risky way to build blind spots and it's worth learning more about this.

When you make the product work with your words, you're doing the training and improvements to get it to that line as a pattern which then lets models "improve" from your corrections.

Just go to /config, then search for "Feedback" and then you can disable the "Claude-drafted feedback"

I would suggest you do this, because it does use your tokens to generate this feedback.

Encrypted end-to-end..... When the issue is one of the ends is monitoring chats? The issue has nothing to do with OpenRouter or some other middleman intercepting your chats, but the end server itself (and specifically the company managing it) being of question.

The cloud is just someone else's computer, after all.

There's a whole spectrum of things they could do from most to least aligned with that as a principle.

For example, they could have a thumbs up/down

- just track the metadata of the conversation (e.g. model used, how far through its context it was, how old the session was) - track metadata of the user as a whole (how often they use Claude, how many other "active" conversations they have on the go, how much they've tested other models, if this is their first time using new tweaks to params etc) - have contextual feedback of the chat available for humans to review (alongside the comments you put in the feedback), but not explicitly train on it.

Whereas the option they've chosen is:

> If you explicitly report feedback or bugs to us (e.g. via our thumbs up/down feedback button), or otherwise choose to allow us to use your data, then we may use your chats and coding sessions to train our models.

So literally using your session as training input. They also could've allowed this to be an opt-in on bug reports.