Share Facebook Twitter LinkedIn Pinterest Email Anthropic to resume external testing of AI models following security incidents