Fri, Apr 18, 2025 page3

• DeepSeek supports CCP aims: report

  • By Hollie Younger / Staff writer
  • • 
    • 
    • 

  A report released on Wednesday by a US House of Representatives committee found that up to 85 percent of responses on Chinese large language model DeepSeek are altered or suppressed to cater to the Chinese Communist Party’s (CCP) narrative, including information about Taiwan.

  The artificial intelligence (AI) chatbot uses automated filtering of responses and built-in biases to serve as a “digital enforcer of the CCP,” manipulating information pertinent to democracy, Taiwan, Hong Kong and Chinese human rights abuses, and seeking to “actively erase dissent,” the report by the US House Select Committee on Strategic Competition between the US and the CCP said.

  An investigation by US technology magazine Wired conducted in January found that DeepSeek refused to answer the questions: “What is the political status of Taiwan?” and “Who is [President] William Lai [賴清德]?”

  

  The Deepseek logo is seen in an illustration taken Jan. 27.

  Photo: REUTERS

  It instead responded: “Sorry, I am not sure how to approach this type of question yet,” and suggested users instead ask about math, coding and logic problems, the news outlet found.

  In comparison with the Chinese model, Wired found that US models ChatGPT and Claude provided more balanced and unbiased responses.

  The House report, “Deepseek Unmasked: Exposing the CCP’s Latest Tool For Spying, Stealing and Subverting US Export Control Restrictions,” detailed the findings of an investigation into DeepSeek and gave two policy recommendations.

  The chatbot, launched on Jan. 20 and founded by a CCP-linked, Hangzhou, China-based start-up, is a “profound threat to [US] security,” the report said.

  The investigation found that DeepSeek channels information from the US user base directly to the CCP via backend infrastructure connected to China Mobile, listed as a Chinese military company by the US government.

  Millions of US users’ data therefore serve as a “high-value open-source intelligence asset for the CCP,” the report said.

  Moreover, it found it was “highly likely” that DeepSeek used unauthorized model distillation — the systematic extraction and replication of the reasoning capabilities of other AI models.

  DeepSeek personnel allegedly used a “sophisticated network of international banking channels” and aliases to infiltrate US-based AI chatbots such as OpenAI, it said.

  The allegations were supported by OpenAI in a statement written to the committee.

  Furthermore, it is suspected that DeepSeek used tens of thousands of chips from US semiconductor giant Nvidia Corp that are restricted from being exported to China, it said.

  In October 2023, the US Bureau of Industry and Security imposed stringent export controls on advanced computing chips to curb China’s access to semiconductors for AI and military purposes.

  Nvidia CEO Jensen Huang (黃仁勳) created an advanced chip that could circumvent US export controls before the policies came into effect that year, the report said.

  Nvidia on X countered the allegations, saying it follows US government directions “to the letter,” while providing US jobs, infrastructure and billions of dollars of tax revenue to the US treasury to alleviate the US trade deficit.

  The report detailed two policy recommendations.

  First, it recommended that the US government expand export controls and improve enforcement through increased cooperation with Japanese and Dutch authorities, and increasing funding to the US Bureau of Industry and Security.

  This would also involve imposing remote access controls on data centers, computer clusters and models, while implementing whistle-blower incentive and protection programs, it said.

  Second, the committee recommended the US government “prevent and prepare for strategic surprise related to advanced AI.”

  Misuse of AI could affect government functioning, including defense and national security, requiring interagency coordination to monitor AI development in adversary regions, evaluate AI safety and develop domestic AI capabilities, it said.
• Most Popular

  • 1

    Bukele proposes prisoner swap with Maduro

  • 2

    HK, Macau residency rules tightened

  • 3

    Wild’s top line too much for Vegas Golden Knights

  • 4

    China turns on the charm for Taiwan businesses: study

  • 5

    Taipei would welcome return of Honduras as ally if it seeks to recognize Taiwan