I'd argue that OpenAI and Anthropic are far more dangerous and I also haven't seen a single convincing argument against that.