How can we send telemetry to help the labs releasing open weights?

I'm the kind of guy who immediately turns off telemetry and error reporting 1st thing when I install a new app. For many apps I even firewall them to prevent phoning home. The only exception: open-source projects. For those, I even go out of my way to check the opt-in since they tend to have it off by default. What strategy can I follow to help companies like Deepseek, Alibaba, GLM, Moonshot, etc (down to the smallest org like Nous Research, ideally), have access to my local prompts, application and tool usage? However, I want to do this without allowing this data to be used by the likes of Anthropic, OpenAI and Google. Some thoughts I had: - Writing a proxy to log all my conversations with coding agents, then periodically sending bullshit summarization requests of the full conversation to the cheapest model on each of their APIs, after opting in to "help improve models." But this doesn't come close to the degree of telemetry companies like Anthropic get from tools like Claude Code. (which even monitors how long it takes you to choose an answer when they give you a multiple choice question) - Thought of switching from Claude Code to Qwen Code when I do local development (currently I use Claude Code for both work and local personal dev): but Qwen Code doesn't even have telemetry that sends to Alibaba. The telemetry is only for your own self-hosted monitoring. Plus this would only benefit Alibaba, I prefer to help all teams. Is there some community project underway to help crowdsource this data, and specifically restricts from using it to train closed models? Like when Mozilla had those crowdsourced ASR and location projects.