To date, frontier AI models have been created exclusively by tech companies. This may change if other companies decide that having an in-house model, trained on their proprietary data, will be sufficiently valuable to warrant training their own frontier model.

The number of companies producing frontier AI models is relevant for AI safety and governance because, all else equal, more actors results in more coordination difficulties. Additionally, there being more actors makes it more likely that one of the actors will be reckless.

—

Thanks to user dschartz, whose question inspired ours.