> Anthropic seems to me to be relatively non-evil, too.

Eh... maybe? We don't yet know the results, but they have been proponents of heavy regulatory interventions since forever. Their plan was basically regulatory capture, where they sell their FUD regarding alignment, "safety" and all that jazz. If they succeed that will be evil, IMO.

The best thing that can happen for us regular users is both healthy competition at the SotA level (which we kinda have, with the big4 labs keeping eachother honest) and support for small open source local models (gemmas, llamas, mistrals, qwens, etc).