What would be surprising is if this happened without any specific training on model experiments and interpretability.

What's surprising though is that I'd expect this type of training to be exclusive to internal model snapshots or at least heavily guardrailed against in public models.

>What could make this wrong: >1. "Understood" means you know every word. It ignores grammar, kanji readings, speaking speed and context.

But at the same time the tax is capped at 15% of the providing company's French revenue or 2% of its worldwide revenue, and realistically speaking, not all companies would pass the tax on to the users, so it could end up being a noticeable but not catastrophic additional expense for both users and companies, if the bill passes at all, that is.