Kagi AI assistant changed to Kimi-2 without my consent

Felensis

I've created two custom AI assistants a few weeks ago which I've configured to use GPT 4o mini. Today I've discovered that both of them were set to Kimi-2 which seems to be a propietary Chinese AI (!). I don't have any trust in a Chinese AI especially if it is propietary. That is a huge no-go in terms of privacy and security to me.

My custom AI assistant shall stay the way like I have set them up and if there are any automatic changes to them then I want to be notified about it. Such changes also should always require my confirmation aka Opt-in.

laiz

Hey! I believe what happened was that upon 4o mini being removed from the Assistant, the custom assistants that were using it defaulted to the option at the top of the list, which in this case happened to be Kimi K2.

I agree this is not ideal. Probably the best approach would be to simply have the custom assistant not work until the user chooses another model when the one they had chosen is retired.

fs1010

It’s an open source model, same as Qwen and DeepSeek and GLM which are hosted in America by third party providers: https://help.kagi.com/kagi/ai/llms-privacy.html

If anything it’s more private and secure than OpenAI’s closed source models, especially with them being sued by the New York Times which is forcing them to keep everything you send to their servers (avoided by using Microsoft Azure, but o3/o3pro/GPT5 are unavailable there and it can fall back to their servers if Azure is at its limit).

Felensis

Alright thanks to both of you for clarification.