DeepSeek Jailbreak Reveals Its Entire System Prompt

Researchers jailbroke the Chinese generative AI DeepSeek to extract its system prompt, raising concerns about model controls and potential transferred knowledge from other providers; concurrently DeepSeek experienced large-scale DDoS attacks and a publicly exposed database (including chat histories, secret keys, and API secrets), while independent tests found the model more biased, toxic, and prone to producing unsafe code and harmful content compared with peers.