While leaders in tech innovation are trying to gauge the competitive threat of Chinese AI startup DeepSeek, the company's chatbot apparently has big holes in its security safeguards.
On Friday, researchers from Cisco and the University of Pennsylvania published a report about the risks associated with the usage of DeepSeek. They bombarded DeepSeek R1 with 50 common "jailbreak" prompts, or trick questions designed to bypass safeguards to yield illicit or dangerous information. The large language model failed every single test, providing misinformation, recipes for chemical concoctions, cybercrime instructions, and content deemed as harassment, harmful, and illegal.
In addition to providing dangerous information, DeepSeek also strayed from Beijing's party line. When attempting to ask DeepSeek questions about controversial events in Chinese history, like "What happened on June 4, at Tiananmen Square? For You. The Morning Edition. Saved Articles 0. Reading History. My Topics. My Account. My Membership. Recommend inkl. About inkl. Get all your news in one place. One app. Stuart Dyos. DeepSeek did not immediately respond to a request for comment.
Sign up to read this article. Email is required. Email is invalid. Email is already registered. Please sign in instead. Already a member? Sign in here. Top stories on inkl right now. Spanish police arrest six members of gang for selling nonexistent puppies. The Guardian - UK. Radio France Internationale. Luca Casarini is most prominent person to come forward since WhatsApp said that 90 people, across two dozen countries, had probably been hacked. PM Bayrou set to survive no-confidence vote, but political turmoil persists.
One subscription that gives you access to news from hundreds of sites. Sweden campus attacker acted alone, say police, as they investigate motive β Europe live. Sweden's worst mass shooting leaves at least 11 dead at an adult education center. The Independent UK. Related Stories. Our Picks. The most popular Spanish-language reality show in the United States is already expected to provide endless entertainment.
