DeepSeek’s Safety Guardrails Failed Every Test Researchers Threw at Its AI Chatbot
Security researchers tested 50 well-known jailbreaks against DeepSeek’s popular new AI chatbot. It didn’t stop a single one.
Read more at Wired
Topics
-
DeepSeek says it built its chatbot cheap. What does that mean for AI's energy needs and the climate?
Chinese artificial intelligence startup DeepSeek stunned markets and AI experts with its claim that it built its immensely popular chatbot at a fraction of the cost of those made by American tech ...ABC News - 2d -
DeepSeek Says It Built its Chatbot Cheap. What Does That Mean for AI’s Energy Needs and the Climate?
If artificial intelligence can truly run more efficiently, the power it needs might be less than experts assume.Inc. - 2d -
Chinese AI chatbot DeepSeek censors itself in realtime, users report
Depending on version downloaded, app approaches its answers with preamble of reasoning that it then erases. We tried out DeepSeek. It worked well, until we asked it about Tiananmen Square and ...The Guardian - 3d -
What will Washington do about Chinese startup DeepSeek and its AI chatbot?
As the TikTok ban hangs in the air, the US investigates the security implications of another mega-popular Chinese app. Who is behind DeepSeek and how did it achieve its AI ‘Sputnik moment’? ...The Guardian - 3d -
Former OpenAI safety researcher brands pace of AI development ‘terrifying’
Steven Adler expresses concern industry taking ‘very risky gamble’ and raises doubts about future of humanity. A former safety researcher at OpenAI says he is “pretty terrified” about the pace of ...The Guardian - 3d -
Trump says China’s DeepSeek AI chatbot is a ‘wake-up call’
Emergence of cheaper Chinese rival has wiped $1tn off the value of leading US tech companies. What is DeepSeek and why did US tech stocks fall? Business live – latest updates Donald Trump ...The Guardian - 3d -
DeepSeek hit with ‘large-scale’ cyber-attack after AI chatbot tops app stores
Attack forces Chinese company to temporarily limit registrations as app becomes highest rated free app in US. DeepSeek said its newly popular app was hit with a cyber-attack on Monday, which forced ...The Guardian - 4d -
Trump AI Order Raises Questions About its Goals and Safety Focus
American AI should be free from ‘ideological bias,’ a new Trump executive order demands, but what does this mean?Inc. - Jan. 24 -
Nvidia tackles agentic AI safety and security with new NeMo Guardrails NIMs
Nvidia is optimizing AI safety controls for multi-agent systems with minimal performance impact through new NeMo Guardrails microservices.VentureBeat - Jan. 16
More from Wired
-
Here’s How DeepSeek Censorship Actually Works—and How to Get Around It
A WIRED investigation shows that the popular Chinese AI model is censored on both the application and training level.Wired - 2h -
OpenAI’s o3-Mini Is a Leaner AI Model That Keeps Pace With DeepSeek
On the heels of DeepSeek R1, the latest model from OpenAI promises more advanced capabilities at a cheaper price.Wired - 2h -
Kash Patel Says He Never Promoted QAnon. Here Are All The Times He Did
Kash Patel helped keep the QAnon movement alive on Truth Social, interacting with Q accounts, appearing on QAnon podcasts, and signing his children’s books with a QAnon catchphrase.Wired - 3h -
RFK Jr.’s Senate Testimony Is Haunted by His Track Record
During his confirmation hearings this week, Robert F. Kennedy Jr. said he would promote vaccines as HHS secretary, despite a long history of promoting anti-vax positions.Wired - 4h -
DeepSeek Has Gotten OpenAI Fired Up
After a Chinese-startup roiled the industry, OpenAI readies a response—ahead of schedule.Wired - 6h
More in Tech
-
Beyond benchmarks: How DeepSeek-R1 and o1 perform on real-world tasks
o1 is slightly better at reasoning, but DeepSeek-R1 provides much more details about its reasoning, which is very useful to the user.VentureBeat - 2h -
President Trump to meet with Nvidia CEO Jensen Huang at the White House
President Donald Trump is meeting Friday with Nvidia CEO Jensen Huang, whose company designs and supplies the advanced computer chips that play an integral role in developing artificial intelligenceABC News - 2h -
Here’s How DeepSeek Censorship Actually Works—and How to Get Around It
A WIRED investigation shows that the popular Chinese AI model is censored on both the application and training level.Wired - 2h -
More rats infesting cities amid warmer temperatures, scientists say
A new study finds that rat infestation in many of the world’s cities appears to be soaring, especially in WashingtonABC News - 2h -
OpenAI’s o3-Mini Is a Leaner AI Model That Keeps Pace With DeepSeek
On the heels of DeepSeek R1, the latest model from OpenAI promises more advanced capabilities at a cheaper price.Wired - 2h