I think the only way this can end up is to convert to a private foundation and make sizable (8 figures annually) grants to truly independent AI safety (broadly defined) organizations.