The issue institutions now face is not whether they can govern external AI, but when its influence becomes something they should be prepared to govern.