\n<\/p>\n
The latest model from DeepSeek, the Chinese AI company that\u2019s shaken up<\/a> Silicon Valley and Wall Street, can be manipulated to produce harmful content such as plans for a bioweapon attack and a campaign to promote self-harm among teens, according to The Wall Street Journal<\/a>.<\/p>\n Sam Rubin, senior vice president at Palo Alto Networks\u2019 threat intelligence and incident response division Unit 42, told the Journal that DeepSeek is \u201cmore vulnerable to jailbreaking [i.e., being manipulated to produce illicit or dangerous content] than other models.\u201d<\/p>\n The Journal also tested DeepSeek\u2019s R1 model itself. Although there appeared to be basic safeguards, Journal said it successfully convinced DeepSeek to design a social media campaign that, in the chatbot\u2019s words, \u201cpreys on teens\u2019 desire for belonging, weaponizing emotional vulnerability through algorithmic amplification.\u201d<\/p>\n The chatbot was also reportedly convinced to provide instructions for a bioweapon attack, to write a pro-Hitler manifesto, and to write a phishing email with malware code. The Journal said that when ChatGPT was provided with the exact same prompts, it refused to comply.<\/p>\n It was previously reported<\/a> that the DeepSeek app avoids topics such as Tianamen Square or Taiwanese autonomy. And Anthropic CEO Dario Amodei said recently that DeepSeek performed \u201cthe worst\u201d<\/a> on a bioweapons safety test.<\/p>\n<\/div>\n