I think Mark Nelson had succinctly described the problem with OpenAI (more broadly, with the "AI safety" concerns): https://twitter.com/mjntendency/statuses/859189428748791810