We all love freedoms. In some legal justifications the big digital firms get all the freedoms they want; the average Joanna, not so much.
In the EU they have the EU AI Act and GDPR. In the US they have the US CLOUD Act. This means the administration has sanctioned judges in Belgium to have their digitally cancelled. Microsoft has leaked unreacted emails from Dutch Trade officials under the US CLOUD Act.
I am sure we all like to talk about over regulation but sometimes those citizen focused regulations are nice to use from time to time. Like rules of love them, its nice when the rule of law is pro your own privacy as you go about your business.
With AI Labs you may be feeling out of luck to send your tokens to somewhere that likes a bit of working checks-and-balances. Yet you are in luck! Mistral AI in Paris is a foundation AI lab, and it is now hosting its first 3rd party open-weights model in France with Z.ai GLM-5.2 zai-glm-5-2. If you fancy trying out there services they have an opensource python harness mistral-vibe (aka Claude Code or Codex CLI thingymabob).
The one catch is that it does not show zai-glm-5-2 by default. I noticed that it didn't show their formal methods proof solving model called Leanstral 1.5 either; yet it had a feature built to switch to to "lean" mode using Shift+Tab. So I figured out the settings to be able to set glm-5.2 as the default model else a model you can switch to. The details are on my gist Running Z.ai's GLM-5.2 on stock Mistral Vibe — no fork required.
Enjoy!
Top comments (0)