Re:Activating AI Safety Level 3 Protections
by alexdong on Thursday May 22, 2025 @02:16PM
(#44064992)
> To be clear, we have not yet determined whether Claude Opus 4 has definitively passed the Capabilities Threshold that requires ASL-3 protections. Rather, due to continued improvements in CBRN-related knowledge and capabilities, we have determined that clearly ruling out ASL-3 risks is not possible for Claude Opus 4 in the way it was for every previous model, and more detailed study is required to conclusively assess the model’s level of risk. (We have ruled out that Claude Opus 4 needs the ASL-4 Standard, as required by our RSP, and, similarly, we have ruled out that Claude Sonnet 4 needs the ASL-3 Standard.)
Re:Activating AI Safety Level 3 Protections
by potholereseller on Thursday May 22, 2025 @03:21PM
(#44065732)
They claim that their new LLM, unfettered, is so good that they are concerned that it might be able to provide instructions to make things like nuclear weapons. A mediocre physicist could provide similar, or better, instructions, but nobody is trying to shut down physics departments at universities.
This is a sales pitch wrapped in over-compliance and presented as an alert. This is pathetic, but I am somewhat entertained by their desperation; I will be even more entertained when LLM companies start using sex-appeal to sell subscriptions ("women find our latest model weights to be irresistible, so sign up now for our $1000/month plan").
The grip, it begins to slip ...
(I, for one, welcome ...)
This is a sales pitch wrapped in over-compliance and presented as an alert. This is pathetic, but I am somewhat entertained by their desperation; I will be even more entertained when LLM companies start using sex-appeal to sell subscriptions ("women find our latest model weights to be irresistible, so sign up now for our $1000/month plan").