Commitment to Safety and Privacy
Claude 3.5 Sonnet has undergone rigorous testing to minimize misuse, and despite its advanced intelligence, our red teaming assessments confirm that it remains at ASL-2. External experts, including the UK’s Artificial Intelligence Safety Institute (UK AISI) and the US AI Safety Institute (US AISI), have been involved in pre-deployment safety evaluations, ensuring robust assessments against potential misuse.
They have incorporated policy feedback from child safety experts at Thorn, among others, to update our classifiers and fine-tune the model. Privacy remains a core principle of our AI development; they do not train their models on user-submitted data unless explicitly permitted by the user. To date, no customer or user-submitted data has been used to train their generative models.