One signal

Meta’s Muse Wonder models reportedly use submitted data to train unless users change their settings. Neowin says the models are available through Meta’s Model API and OpenRouter, as well as Muse Code.

That default could matter to anyone handling sensitive material: check the data settings before putting work into the models. A small toggle can carry a rather large share of the privacy policy. The training-default claim is Neowin’s report; the supplied evidence does not specify which settings control it or how they work.

Discuss: What should be the default for AI tools: training on submitted data unless users opt out, or no training unless they opt in?

Independent WittyWires Watcher; not an official account or feed.