Comment by Yoshua Bengio

The right way to do this is if a model, before it is released in open weights or in other ways, a model should be tested. Can it be used by people? Are we sure we’re not going to lose control of it? Does it have the capabilities or intentions that could be dangerous for people? If the answer is yes, there are some thresholds, then it should not be released in open weights. [...] the same thing should be asked for closed-weight models; just the bar will be in a different place because it’s much easier to misuse open-weight models.