On 4 July 2024 the World AI Conference and the High-Level Meeting on Global AI Governance opened in Shanghai. The Shanghai Declaration on Global AI Governance was issued the same day, setting out consensus around promoting development, safety governance, openness and inclusiveness, and international cooperation. It is a statement of position: no monitoring mechanism, no specific obligations. What matters is the order in which it places its keywords.
The order of the four clusters
“Promoting development” comes first, and the order is not rhetorical. International AI texts have usually started from risk; this declaration puts development and governance side by side, which accepts that regulation cannot come at the cost of industry. “Safety governance” follows, pointing at controllability. “Openness and inclusiveness” and “international cooperation” close the text, answering fragmentation and market division.
Together the four clusters form a usable framework for external statements, but they leave the hard questions open: who sets the safety standard, how wide openness extends, and on what terms cooperation happens. The declaration supplies shared vocabulary; the operative rules will still be filled in by legislation and enforcement in each jurisdiction.
Why Shanghai
The choice of Shanghai reflects the density of the local industry. Model developers, compute providers, device makers and application companies cluster in the same region, so governance questions land on concrete products more easily. For overseas participants, Shanghai is also a window onto Chinese AI industrial policy.
Industrial policy and international narrative appearing at the same event invites a common misreading: that policy support means lighter regulation. The opposite tends to hold. The larger the industry, the more specific the enforcement of filing, labelling, security assessment and data compliance. The capabilities a company displays at an exhibition are often the same ones regulators are examining.
What exhibition and demonstration mean legally
Booth demonstrations, launches and media interviews are external statements. Claiming that a product is “safe”, “trustworthy” or “controllable” works commercially, but in legal terms it may be treated as a statement of fact. Capability claims that cannot be substantiated become a liability in customer diligence or a regulatory inquiry; absolute wording can also raise advertising-compliance issues.
Demonstrations carry a second layer of risk. A live demo using real user data requires separate authorisation and minimisation. Using information from overseas attendees also involves cross-border transfer. Substituting synthetic data and processing locally is a low-cost arrangement.
Turning the declaration into internal documents
A company does not need a separate compliance programme for a declaration. The workable approach is to distribute it into existing documents: a board-level AI governance policy that states the balance between development and safety; a risk register covering models, data and use scenarios; supplier contracts requiring change notification and limits on data use; an incident-response process that fixes the external communication line.
In external references, explaining how a paragraph of the declaration maps to actual practice is more persuasive than reproducing principles. The declaration is not a standard and does not replace mandatory rules in the EU, the US or China. Its function is to give the dialogue across jurisdictions a common starting point.
For cross-market operations and rule comparison, see our multi-jurisdiction compliance service and EU AI Act service. For data arrangements in demonstrations and operations, see data compliance.
This article is based on the Shanghai Declaration on Global AI Governance issued at the World AI Conference and the High-Level Meeting on Global AI Governance on 4 July 2024, and on public information. It is for reference only and does not constitute legal advice.