I think Apple has really got ahead of this game: early deals to pay for AI training data/content. I need to do some research but I think Anthropic also does this.
After a year of largely using OpenAI APIs, I am now much more into smaller “open” models for I hope the major contributors like Meta/Facebook are following Apple’s lead. Off topic, but: even finding the smaller “open” models much less capable, they capture my imagination and my personal research time.