I actually think unilateral pause by a lab goes on this list. As in OpenAI or Anthropic going “hey everybody, we think it’s getting too dangerous to continue pushing the frontier, so we are pausing development of new frontier models. We’ll keep serving the models we have, but all of our R&D will be towards safety until we’ve solved it.” The hope being that this sends a costly signal to regulators that this is serious and other frontier labs should be stopped as well. It needn’t even be a bad business move—they could say “those other labs keep putting out smarter models, sure, but what good is it if they might get your company in legal trouble by going rogue? Our models are safer.” Or something like that.
I actually think unilateral pause by a lab goes on this list. As in OpenAI or Anthropic going “hey everybody, we think it’s getting too dangerous to continue pushing the frontier, so we are pausing development of new frontier models. We’ll keep serving the models we have, but all of our R&D will be towards safety until we’ve solved it.” The hope being that this sends a costly signal to regulators that this is serious and other frontier labs should be stopped as well. It needn’t even be a bad business move—they could say “those other labs keep putting out smarter models, sure, but what good is it if they might get your company in legal trouble by going rogue? Our models are safer.” Or something like that.