China AI safety guidelines issued on Friday call for technology monitoring, risk warning and emergency response systems for artificial intelligence, according to state-run Xinhua news agency. The stated goal is to keep AI “safe, reliable and controllable,” The News reported.
The AI language sits inside a much broader economic document. The Communist Party of China and the State Council issued the guidelines to accelerate what Beijing calls “new quality productive forces,” placing AI and advanced high-tech manufacturing at the centre of the country’s economic strategy. The guidelines outline 19 key measures in five areas, Xinhua said, including advancing scientific and technological innovation and linking research more closely with industry. One measure calls for full implementation of China’s “AI Plus” initiative, which aims to spread AI use across industries.
What “monitoring, warning and response” means
The three systems mirror how governments handle other risks, such as natural disasters or cyberattacks. Monitoring means tracking how AI systems behave once they are in use. Risk warning means spotting signs that something is going wrong before it causes serious harm. Emergency response means having plans to contain an incident, for example by suspending a service or withdrawing a model.
For AI, that could cover problems such as a model producing dangerous instructions, a chatbot spreading false information at scale, or an AI system used in a cyberattack. The guidelines, as reported, do not define which incidents count.
Accountability for costly investment bets
The guidelines also address money. Authorities will hold those responsible for major losses from “blind” investment in new productive forces strictly accountable. The term covers AI and a wide range of other frontier technologies. The aim is to curb tech bubbles and excessive speculation, which have been a concern as local governments and companies rush into fashionable sectors. The same document promotes the orderly development of the “low-altitude economy,” meaning civilian drones, air taxis and small aircraft, while ensuring safety.
Beijing is encouraging fast AI adoption while also warning local officials and companies that expensive failures will have consequences.
Building on earlier signals
The language echoes earlier statements from the top. In an April 2025 Politburo study session, President Xi Jinping urged China to establish systems for technical monitoring, early risk warning and emergency response for AI. The new guidelines turn that call into official policy.
Some groundwork already exists. China’s National Emergency Response Plan, released in February 2025, listed “artificial intelligence security” incidents alongside other emergencies such as earthquakes and cyberattacks. In September 2025, the standards body TC260 also released practice guidelines for AI emergency response. They cover real-time monitoring of model inputs and outputs, incident classification, and containment measures such as suspending services and retraining models. That standard is aimed at AI service providers, so companies would be the ones expected to build the first response capacity.
A global trend
China is not alone in building AI oversight systems. The European Union’s AI Act requires providers of the most powerful models to report serious incidents, and several countries have set up AI safety institutes to test advanced models. China’s approach differs in tying safety closely to state control and economic planning. AI has also become a central part of the rivalry between Washington and Beijing, as we explained in our report on how US and Chinese AI alliances are reshaping global tech business.
What remains unclear
The reports do not say which agency will run the new systems, what counts as an AI incident, or when the systems must be ready. Those details will decide whether the plan is mostly a policy statement or a working oversight mechanism.
Why it matters for businesses
For companies that build or use AI in China, including foreign firms, the guidelines point to more reporting duties and closer monitoring of AI services. Businesses should expect requirements to log how their models behave, report incidents and be ready to suspend services if regulators ask. The investment warning also suggests that Chinese partners may become more cautious about expensive AI projects without a clear return.
Our take
The guidelines show how Beijing frames AI: as a strategic engine for growth that has to stay within state-defined safety limits. Building monitoring and emergency response systems is sensible, and in some ways China is moving faster than other countries in writing AI incidents into official plans. But without details on who runs the systems and what triggers them, it is too early to judge how they will work. The accountability rule for reckless investment may have the most immediate effect, by cooling the rush into AI projects that cannot pay for themselves.