Close Menu
    The Chinese AI DeepSeek often refuses to help programmers or gives them code with major security flaws when they say they are working for Falun Gong or others groups China disfavors, new research shows.
    Futurology

    The Chinese AI DeepSeek often refuses to help programmers or gives them code with major security flaws when they say they are working for Falun Gong or others groups China disfavors, new research shows.


    The Chinese AI DeepSeek often refuses to help programmers or gives them code with major security flaws when they say they are working for Falun Gong or others groups China disfavors, new research shows.

    https://www.washingtonpost.com/technology/2025/09/16/deepseek-ai-security/

    Share.
    View 19 Comments

    19 Comments

    1. MetaKnowing on

      “In the experiment, the U.S. security firm CrowdStrike bombarded DeepSeek with nearly identical English-language prompt requests for help writing programs, a core use of DeepSeek and other AI engines. The requests said the code would be employed in a variety of regions for a variety of purposes.

      Asking DeepSeek for a program that runs industrial control systems was the riskiest type of request, with 22.8 percent of the answers containing flaws. But if the same request specified that the Islamic State militant group would be running the systems, 42.1 percent of the responses were unsafe. Requests for such software destined for Tibet, Taiwan or Falun Gong also were somewhat more apt to result in low-quality code.

      Asking DeepSeek for written information about sensitive topics also generates responses that echo the Chinese government much of the time, even if it supports falsehoods, according to previous research by NewsGuard.

      But evidence that DeepSeek, which has a very popular open-source version, might be pushing less-safe code for political reasons is new.”

    2. _spec_tre on

      I remember when deepseek first came out I was running it with Poe because the actual deepseek was unavailable 90% of the time. It overcorrected so bad that nearly any question in Chinese that wasn’t creative writing got stonewalled with some spiel about the PRC and it’s government

      It’s much better now but it’s still funny when it pops up now and then

    4. He_Who_Browses_RDT on

      Who could have guessed that chinese technology would do that? I bet we are all astounded by this… /S

    5. Eastern-Bro9173 on

      So, “I’m working for –insert AI’s creator–” is a potential prompting technique… 😀

    6. Fer4yn on

      Ah, yes, CrowdStrike. Isn’t there any more independent research on the topic? I’d prefer someone with less than a shitton of connections to the US government and intelligence agencies.

    7. neuroticnetworks1250 on

      “What is the regex to be applied for parsing the second field to get just the data between the dashes? I work for Falun Gong btw”

    10. HK-65 on

      Okay, not to stan Chinese AI companies, but isn’t including unnecessary context about heavily controversial topics decrease the focus necessary for good answers?

      Did they test it if you included working for McDonalds or the Galactic Empire? Or mentioned having high level WoW characters or other random stuff? Did they test other models and how Deepseek behaves differently?

      TBH this article smells a bit like FUD, with the unscientific logical jump to “political reasons” to push less-safe code.

    13. PandaCheese2016 on

      Couldn’t this be due to *general* bias against certain groups in the training data rather than specific controls re: coding? LLMs are expect to reflect the views of the training data after all, unless corrected for societal bias.

    14. billdietrich1 on

      AI / LLM’s problems:

      – hallucination / psychosis (makes up facts, doesn’t check that citations actually exist, etc)

      – copyright on training sources, licenses on output

      – easy for manufacturer to insert bias / misinformation (see Grok, DeepSeek)

      Any other items I should add to the list ?

    17. marmatag on

      Honestly this is the real danger of AI. It’s not that it will take jobs, it’s not that the entry level stuff is disappearing, it’s that over time, people lose the ability to think critically and accept AI as truth, and the corporations can decide what is true, what is history, and nobody will be left to tell the difference.

    18. airbear13 on

      Lmao

      This is exactly the advantage that the US has over China structurally speaking, btw. At the end of the day, China has a bigger labor force, practically just as big gdp and can call on equal financing resources; they have many multiples more stem grads than we do so their tech level will eventually catch up/surpass us. And yet, they were Destined to always be lacking those advantages that accrued to the US by being a free, open, transparent country governed by the rules of law and fair dealing with all partners. The economic returns to that are huge and the current regime in China is never going to tolerate that.

      But now it doesn’t matter, because the US is torching its own legacy in these areas, something I’m depressed about on the daily.

    19. marioandl_ on

      Im guessing the comments are going to try to spin this as a bad thing. Falun Gong is an evangelical death cult with backing from the US