The latest model from Deepseek, the Chinese Ki company, that is shaken Silicon Valley and Wall Street can be manipulated to create harmful content such as plans for a biowaffe attack and a campaign to promote self-harm between teenagers. According to the Wall Street Journal.
Sam Rubin, Senior Vice President at Palo Alto Networks’ Threat Intelligence and Incident Response Division Unit 42, told the journal that Deepseek was “more susceptible to jailbreaking (i.e. manipulated in order to produce illegal or dangerous content) than other models.”
The Journal also tested the R1 model from Deepseek itself. Although there seemed to be fundamental protective measures, Journal Deepseek successfully designed a social media campaign that promoted in the words of the chatbot “the desire of the youngsters and emotional Vulnerability due to algorithmic reinforcement weapons ”.
According to reports, the chatbot was also convinced that they would give instructions for a biowaffe attack, to write a Pro Hitler manifesto and to write a phishing email with malware code. The journal said when Chatgpt made the same input requests available, it refused to fulfill.
It was previously reported That the deepseek app avoids topics such as the tianamen or Taiwanese autonomy. And the anthropic CEO Dario Amodei recently said that Deepseek played “The worst” On a bioweapons security test.
