US House Committee finds 85% of DeepSeek AI chatbot replies filtered to CCP narrative
The chatbot employs automated response filtering and inherent biases to function as a "digital enforcer of the CCP," distorting information related to democracy, Taiwan, Hong Kong, and Chinese human rights violations, in compliance with Chinese laws and aiming to "actively erase dissent," the Taipei Times reported.
ANI
Apr 17, 2025 13:12 IST
Taipei [Taiwan], April 17 (ANI): A report published by the US House Select Committee on the Chinese Communist Party indicated that approximately 85 per cent of the replies generated by the Chinese AI chatbot DeepSeek are modified or suppressed to align with the CCP's narrative, as reported by Taipei Times.
The chatbot employs automated response filtering and inherent biases to function as a "digital enforcer of the CCP," distorting information related to democracy, Taiwan, Hong Kong, and Chinese human rights violations, in compliance with Chinese laws and aiming to "actively erase dissent," the Taipei Times reported.
According to the committee, DeepSeek, which was launched on January 20 and created by a CCP-associated startup in Hangzhou whose controlling shareholder is Liang Wenfeng, a co-founder of the quantitative hedge fund High-Flyer, poses a "profound threat to [US] security," as cited by Taipei Times.
The investigation revealed that DeepSeek transmits information from its US user base directly to the CCP through backend infrastructure linked to China Mobile, which the US government has designated as a Chinese military entity.
Thus, the data of millions of US users is classified as a "high-value open-source intelligence asset for the CCP," according to the committee. Additionally, it determined that it was "highly likely" that DeepSeek utilised unauthorised model distillation, which involves the systematic extraction and duplication of the reasoning abilities of existing AI models.
Personnel from DeepSeek allegedly employed a "sophisticated network of international banking channels" and aliases to infiltrate US-based AI chatbots like OpenAI, it reported. These accusations were corroborated by OpenAI in a statement addressed to the Select Committee.
Moreover, there are suspicions that DeepSeek acquired tens of thousands of chips from US semiconductor leader Nvidia, which are currently prohibited from being exported to the PRC, according to the Taipei Times.
The US Department of Commerce is presently investigating if DeepSeek unlawfully imported export-controlled advanced Nvidia chips via Singapore, an intermediary nation with less stringent export regulations, Taipei Times added. (ANI)
Get the App
