Chinese govt fears open-source AI could undermine information control

New Delhi: Even as China presents itself as a global champion of low-cost, open-source artificial intelligence, the Chinese Communist Party is increasingly concerned that advanced artificial intelligence could undermine its control over information and national security, a new report has said.

A report from The New York Times said party officials fear advanced AI could be exploited by hackers for cyberattacks, scams and terrorism bypassing censorship. The Chinese administration even fears that advanced AI could assist in the development of dangerous biological agents.

Chinese experts reportedly warned about deepfakes, data breaches and AI data poisoning at a recent Cybersecurity Conference in Beijing and that it could spread information that challenges the Communist Party’s version of events, the publication said.

China's ministry of state security had reportedly indicated that hostile forces could exploit AI to spread information about domestic protests, crackdown on Muslim ethnic groups, or Taiwan contrary to official narratives.

"AI data poisoning reportedly manipulates training data to influence AI responses, reportedly raising concerns that chatbots could generate politically sensitive or anti-government content," the report said.

Beijing has promoted low‑cost, open‑source AI as an AI outreach strategy to developing countries similar to its Belt and Road initiative.

Unlike closed AI systems, open-source models can be downloaded, modified and used by anyone, even after safety protections are removed.

As AI grows in advancement, Beijing faces a dilemma of whether it should promote open technology to lead the global race or tighten control to protect the Communist Party’s hold on power, the media house said.

The report said the administration has blocked a Singapore‑based AI acquisition on national security grounds in April and held meetings with leading domestic AI firms to discuss restricting foreign access to top models.

ABOUT THE AUTHOR

IANS

[email protected]