The idea of censoring AI models to be "safe" needs to die. They are tools and such the exact same usage can fundamentally be used for both good and bad. There is no way to differentiate between usage of an AI model for security research or malicious attacks, fictional story telling or advice for a real crime, creative usage in parodies or cat fishing scams.