Authoritarian Innovation
Executive Summary
China’s use of generative AI for information manipulation has drawn growing attention worldwide, but from Beijing’s perspective, the technology is also a risk vector that could weaken its own control over the information environment. Based on qualitative review of publicly available Chinese sources published between 2013 and 2026, this report shows that Beijing is building political control into generative AI systems through an increasingly comprehensive governance mechanism, which enables information control to operate at the technical layer even without deliberate information operations.
Within China’s broader AI safety and security framework, content that deviates from Party narratives or challenges the Party’s ideological worldview is treated as a content security risk. This has driven efforts to embed control into generative AI systems before content is generated.
China’s policy response centres on an expanding system of industrial standards across training data, data annotation, model behaviour, application testing, and content traceability. These standards translate high-level political requirements into implementable technical procedures, ensuring that information control is built into the process before content is generated.
Private technology firms are incorporated into China’s AI governance system, working with state actors to embed the Party’s political requirements into technical standards and model behaviour. This role follows the early-2020s reconfiguration of state–private sector dynamics, when China’s private sector was disciplined and incentivised to align its business and technical capacity with the Party’s political objectives.
As Chinese generative AI models gain traction abroad, Party-aligned content controls embedded in these systems are travelling beyond China’s domestic information space. Without relying on overt information operations, their diffusion creates a less visible form of influence than conventional information manipulation, with risks deepening as foreign developers, platforms, and applications become dependent on Chinese models.
With China seeking a greater role in global AI governance and standard-setting, international regulators and developers should recognise that the country’s language of AI safety and security often carries its own political and ideological requirements. Overlooking this political context risks allowing authoritarian approaches to content control to shape the rules of global AI governance.