Local AI
Models running on hardware you own.
Nothing leaves your machine, there is no per-message cost, and it keeps working if a provider changes their terms. The trade-off is that you need hardware capable of running it, and smaller local models are less capable than the big cloud ones.