Anthropic's safety test results showed that DeepSeek AI does not block harmful prompts, even offering critical bioweapons ...