China releases three datasets for AI training and security research

Source announcement: 2026-09-15. Prepared with AI assistance from the linked primary sources. Company and government claims are attributed to their issuers. How we use AI.

Three AI-related datasets were released at a cybersecurity forum in Jinan on 15 September 2026, according to an announcement published by China’s Cyberspace Administration the following day.

The Chinese Internet Basic Corpus 4.0 contains 120GB of language material. Cybersecurity Dataset 1.0 comprises five collections, including evaluation benchmarks, attack-and-defence data, alerts, anomalous traffic and vulnerability analysis. A separate generated-content detection dataset contains 100,000 text items and 100,000 images.

The datasets were developed by cybersecurity organisations and research partners. They are listed on the Chinese Internet Corpus Resource Platform, reached through the Cyber Security Association of China website. Users must register and complete verification before downloading data or contacting providers to obtain it.

See something we should correct? Contact the newsroom.