ELON MUSK： Why competitors should review each other's AI models before release？

I would recommend that we at least have some sort of informal weekly or biweekly call and that there's maybe a week or two weeks of early access by competitors. This is why I think the incentives work， competing companies are likely to understand if something is a security risk and they're not going to be shy about highlighting that their competitors' models should be delayed in release. The incentive structure does work quite well there. It's similar to how the Motion Picture Association works as an industry group that decides how movies should be rated. That seems to have worked quite well. You really just have to ask the government to intervene if a company is doing something very dangerous and refuses to take action to reduce the danger， which is literally what just occurred with Mythos after Amazon highlighted the cybersecurity risks and called the White House.