Recent warnings from American AI executives have created the impression that the world is only beginning to grapple with the dangers of advanced artificial intelligence. Dario Amodei, chief executive of Anthropic, has urged governments to intervene and slow the development of increasingly powerful AI models. His proposal starts with independent evaluators embedded inside AI companies, moves to regulation, and culminates in government-backed coordination among democratic states, ultimately seeking global agreements with China on testing, dangerous applications, and the pace of AI development.
Yet there is a China-shaped hole in this narrative. The Asian superpower is often portrayed as the difficult final participant in a governance system that the US and its allies must first create. In reality, China has already established an extensive system of AI rules, assessment, filing, technical standards, and government oversight. Here’s how China’s AI governance has developed in stages alongside the technology itself.
From algorithms to deepfakes
Beijing’s 2022 rules on algorithmic recommendation services apply to technologies that select, rank, filter, or recommend online content to users. They compel companies using algorithms to examine how they work, protect users’ personal information, give users certain controls, and introduce safeguards for workers, consumers, children, and older people.
China’s deep synthesis provisions, effective from January 2023, extended regulation to synthetic (or AI-generated) text, images, audio, and video. AI companies must authenticate users, protect data, keep relevant records, assess their systems, and identify synthetic material that could be mistaken for authentic content.
The interim measures for generative AI services, introduced in August 2023, added rules for public-facing generative AI that can influence public opinion or persuade users in other ways. These measures cover the legality and quality of training data, users’ personal information, intellectual property, and discrimination based on protected characteristics. The rules also address transparency by AI companies and the accuracy and reliability of products. Under the rules, companies must address problems found in their services, undergo security assessments to examine risks and compliance with Chinese legislation, and file information about their algorithms with Chinese authorities.
These mechanisms are not identical to the independent evaluators proposed by Amodei, but they perform some of the same functions.
Frontier risks
Beijing’s latest set of measures is the AI safety governance framework 3.0, published in September 2026. It sets out a broad approach to identifying, grading, and responding to risks throughout the lifecycle of AI systems. Building on earlier versions, the framework covers AI model and data risks, vulnerabilities in open source models, AI’s effects on employment and society, and increasingly autonomous systems. It includes a dedicated mechanism for managing risks associated with rapidly developing AI agents—systems that carry out tasks in response to user instructions.
The framework addresses some of the same dangers now generating alarm in Silicon Valley, including highly autonomous systems, the misuse of chemical, biological, and nuclear technologies, and the possibility of humans losing control of advanced AI. It proposes continual safety evaluations, risk-based safeguards, humans in charge, and emergency intervention mechanisms. An independent analysis by the US Carnegie Endowment for International Peace think tank has described the framework as China’s attempt to “advance frontier safety without the need for heavy-handed regulatory interventions.”
Legal consequences
None of this means that Chinese AI rules should be transplanted elsewhere. China’s institutions, regulatory traditions, and technological priorities are different. But many functions now presented as urgent aspirations in the US—evaluation, accountability for AI companies, risk classification, regulatory filing, technical standards, and government oversight—are already embedded in the governance architecture of the world’s other AI superpower.
The US is free to conclude that a market-led system best serves its economic and strategic interests. The Trump administration has favored a “minimally burdensome” national approach to AI and has sought to limit more demanding state-level regulation. But a market-led model must not mean privatizing the rewards of AI while exporting its risks. US companies may capture the commercial gains, while governments elsewhere are left to investigate security incidents, protect public systems, and absorb the wider costs when those technologies fail. The OpenAI agent that gained unauthorized access to Australia’s state healthcare service shows how AI risks are being exported.
If Washington rejects extensive controls before systems are released, it needs a credible legal regime for assessing who is liable for damage caused by those AI models. Otherwise, countries dependent on foreign AI systems will face increasingly difficult choices about whose technology to trust, and on what terms. As treating AI risks like a pathogen suggests, international cooperation is essential.
Silicon Valley’s governing ethos has long been to “move fast and break things.” That was always a questionable principle for technologies woven deeply into public life. It is untenable for companies whose own leaders now warn that their products could cause catastrophic harm. Beijing has already surrounded its AI sector with public obligations, regulatory oversight, and an evolving framework of risk assessment. A global debate that treats such governance as a future aspiration should at least recognize that one country is already ahead.
