Xi sets people-centric AI governance frame at Shanghai conference

The Cyberspace Administration of China published an account on 17 July of President Xi Jinping's opening address to the 2026 World AI Conference and High-level Meeting on Global AI Governance in Shanghai, in which Xi framed AI governance as a system-level political priority rather than a new instrument directed at model providers or deployers.

Xi said countries should uphold people-centric AI principles and a pro-social AI approach, and work together to build a just global AI governance framework. He said AI should become 'an important driving force' for shared prosperity and common security, and that countries should 'join hands to build a fair and reasonable global artificial intelligence governance system.'

The CAC text announces no new measure, implementation date or compliance procedure, and should be read as a political direction-setting statement rather than an operative rule. What it supplies is a high-level vocabulary that compliance and public-policy teams will need to map onto existing Chinese controls, particularly where providers, platforms and deployers already log filings, review outputs and register products under current measures such as the Interim Measures for the Management of Generative Artificial Intelligence Services (CAC, effective 15 August 2023, https://www.cac.gov.cn/2023-07/13/c_1690898327029107.htm). The consultation has not disclosed any affected-licensee count or next filing window.

The foundation-model versus deployed-system distinction remains live in the China file even though the speech does not use that language. Xi's remarks describe a global governance framework in principles terms, while China's operative regime has tended to land through binding administrative measures with content-governance overlays. A provider building or adapting a model, a platform distributing it and a business deploying it into a regulated user journey face different operational questions regardless of where the political messaging sits.

The near-term task for product counsel is translation. Teams should review whether internal policy language, model-card disclosure calibration and governance documentation already align with the people-centric and pro-social AI language used in the leadership statement, particularly where those documents are shown to Chinese regulators, enterprise buyers or state-linked partners.

The open question is whether this political framing is followed by a CAC or State Council instrument that classifies affected systems, updates filing expectations or sharpens review criteria. The speech aligns with China's established pattern — broad governance language at the top, then binding implementation through specific administrative rules — and no such instrument has been announced.