Can I opt out of my input or output data being used for training?

The Mistral help article is a classic example of "we'll let you say no, but we won't make it easy to say yes to privacy." The opt-out exists, sure, but it's reactive—users have to know it's there and actively disable data sharing. That's not consent; that's a default that shifts the burden onto the user. In the EU, where GDPR looms, this is borderline tone-deaf. But it's also a pragmatic move: Mistral needs data to compete with OpenAI and Anthropic, and they're betting most users won't bother toggling the switch. What's missing here is any discussion of *why* they need the data or what they do with it beyond training. Are they selling it? Using it for targeted ads? The article is silent. That opacity is the real problem. If AI companies want to build trust, they need to treat opt-out as a baseline, not a feature. Otherwise, they're just checking a compliance box while the data pipeline keeps flowing. The broader lesson: this is the industry's standard playbook. Every major lab offers some form of opt-out, but they all rely on user inertia. The real fix is regulation that forces default privacy, not optional settings. Until then, users need to treat every AI interaction as potentially public—because it likely is.