Tashkent, Uzbekistan (UzDaily.uz) — Xinhua News Agency has launched a Global Media Initiative to jointly develop standards for the use of artificial intelligence in journalism and published a Framework Document on Global Media Cooperation in AI at the 2026 World Media Summit.
The initiative aims to promote the exchange of experience and cooperation in the responsible and standardised use of AI in journalism, according to materials from the 2026 World Media Summit.
Its main priorities include sharing practices, joint testing of solutions, collaborative research and professional skills development. The initiative is intended to help partners contribute to collective efforts by drawing on their respective strengths.
The first group of partners comprises 11 Chinese and 11 international organisations.
Chinese participants include People's Daily, Xinhua News Agency, China Media Group, Peking University, Tsinghua University, Renmin University of China, the China Copyright Protection Centre, the China Academy of Information and Communications Technology, the Institute of Automation of the Chinese Academy of Sciences, the China National Institute of Standardization and the China Association of Printing Technology.
International partners include TASS, the Vietnam News Agency, Rossiyskaya Gazeta, the Union of OIC News Agencies, Qatar News Agency, Independent Online (South Africa), La Nueva Televisión del Sur, the Malaysian National News Agency, Agencia Peruana de Noticias Andina, the Uganda Broadcasting Corporation and Tunis Afrique Presse.
The framework document draws on Xinhua's practical experience in developing a system of standards for AI use in the news industry. It is intended to facilitate the exchange of risk management practices and experience in applying AI to areas such as data governance, model deployment and content security. It also seeks to establish communication channels between media organisations worldwide. The document is designed as a voluntary reference framework.
Its central principle is that AI should remain an auxiliary tool, while journalists retain ultimate responsibility for news content.
Media organisations are encouraged to uphold the journalistic ethics of truthfulness, combat AI hallucinations and disinformation, strengthen safeguards for accuracy, and pay attention to testing and evaluating model capabilities. The document also emphasises respect for cultural diversity and consideration of the different circumstances and levels of development of media organisations. Its goal is to help build a reliable and diverse global information environment.
The practical framework covers three main areas.
The first is content labelling and provenance verification. Materials substantially generated, rewritten or synthesised using AI should carry visible labels and embedded metadata.
The labels should inform audiences that AI was used, while the metadata should record the extent of AI involvement, the model version and the time of creation.
For high-risk synthetic content, more tamper-resistant safeguards, including digital watermarks, are envisaged where appropriate. The framework also calls for records to be maintained throughout every stage of AI-assisted production, from prompts and source materials to edits by editors and verification personnel. Labels should remain attached when content is republished.
The second area is AI model evaluation. Xinhua's standards system for the news industry combines automated testing with professional journalistic assessment, including checks of factual accuracy and logical completeness. Participants are encouraged to share evaluation tools and test cases that have been validated in practical use.
Evaluations should be objective and impartial, results reproducible and reports standardised.
The third area is security and risk management. The document provides for a closed-loop risk management system covering pre-deployment assessment, access controls and logging during operation, and incident response after deployment.
Before implementation, systems should be tested for risks including factual hallucinations, malicious prompt injection, deepfakes and algorithmic bias. Clear operating boundaries and access rights should be established for AI agents and automated news systems, alongside mechanisms for human intervention and emergency shutdown.
AI-generated results may be used only as drafts or reference materials. All content intended for publication must undergo professional fact-checking against multiple sources.
The use of AI to fabricate quotations or details of events is prohibited. Simulated or reconstructed multimedia materials must be clearly labelled and must not be presented as authentic recordings from the scene. Personal data used in AI training and processing materials should be anonymised, while access to confidential materials must be strictly restricted.
Participants may establish a working group and an open review mechanism following consultations to ensure the document is regularly updated and improved. They are also encouraged to share practical experience from media operations and conduct joint case reviews.
Participation formats, including practical implementation, case sharing, joint model evaluation and thematic discussions, are voluntary.
The document is non-binding and voluntary. Participants may adapt it to their local legal frameworks, and it imposes no obligations.
Media organisations and specialised institutions worldwide have been invited to join the initiative. The document is expected to be updated periodically to reflect developments in AI technology.
