Leaked documents detail Chinese AI surveillance of Taiwan chip sector

Dark Reading reports that leaked documents describe a Chinese government programme that uses AI-powered big‑data mining to monitor overseas ethnic-Chinese communities and identify individuals who could be useful for influence operations. The effort is run from a facility described as the Overseas Chinese Affairs Big Data Platform, headquartered in a 500-square‑metre lab at Jinan University in Guangzhou.

The system aggregates open‑source information from social media and other public sources, translates foreign language content into Chinese, and feeds results into machine‑learning processes to produce reports for CCP leadership. The project is said to combine OSINT with a multi‑tier processing chain, where scraped data is quarantined abroad via proxy networks before being returned to China for analysis.

The reporting identifies a focus on high‑profile figures in the semiconductor sector, including executives of the Taiwanese AI firm Skymizer, leading researchers at a top Taiwanese university, and an AI centre director at Microsoft.

The leaked material, described by NetAskari to Dark Reading and Sky News UK, outlines strategies to build relationships through non‑official organisations and to shape media narratives, with the aim of influencing influential individuals and potentially recruiting assets for Beijing’s purposes.

The platform’s operational details include open‑source scraping from Instagram, Facebook, X, Weibo and WeChat, with scrapers operated in foreign locations and data shielded behind rotating proxy networks to avoid tracing the collection system. Some documents suggest outputs range from public relations trends to targeted outreach plans, while acknowledging uncertainty about whether newer large language models have been integrated into the system since references date from 2021.