Comment by Steven Adler

There is no permanent way to apply safety limitations to prevent users from obtaining help from the model with regard to bioweapons-related tasks. Verified source (2025)