Same Words,
Different Meanings

A shared vocabulary around AI governance may not necessarily imply shared commitments. This article examines divergence in how a set of terms in AI governance are used in China and in the US, the UK, and the EU. This article traces key Chinese language terms’ operational meanings across Chinese official sources pertaining to AI strategy and governance — including within leadership speeches, regulations, technical standards, and state-affiliated strategic analyses, and examines comparable terms’ prominent usages in US, UK, and EU-level official policy discourse. The analysis focuses primarily on how these concepts are understood at the official level, because misunderstandings in diplomatic engagement could lead to false consensus, where policymakers mistake shared terminology for shared understanding even when the underlying concepts reflect different policy objectives. The research aims to support China-facing diplomacy, including Track I engagement.

The analysis is based on sources reviewed through July 2026.

Headline findings

• The same English translation can obscure multiple operational meanings in Chinese AI governance. For example, while ambiguity can also exist in English-language discourse, “loss of control” retains a stable core meaning around human control over AI systems. Chinese official usage, on the other hand, applies it to at least four distinct control relationships, while safety/security (安全) spans five registers. Hence, a single term may refer to different risks or objects of control, often within the same speech or policy text.
• Technical convergence coexists with political divergence. Chinese technical standards identify many risks familiar to frontier AI governance in the US, the UK, and the EU, but place them within wider national security imperatives. Some terms are operationalised through regulatory instruments that prioritise domestic content compliance and regulatory oversight, rather than frontier-risk evaluation or public transparency.
• Operational meaning heavily depends on institution and context. Interpreting the operational meaning of a term requires examining the publishing institution’s background and reading the term against the surrounding text.
• Shared language should not be treated as evidence of shared concepts. When interpreting Chinese policy texts or engaging Chinese counterparts, policymakers should identify what is being controlled or protected, and verify whether the term refers to a technical property, a domestic governance capacity, or a broader geopolitical argument.

The full report features two deep-dives into the concepts of loss of control (失控) and safety/security (安全), alongside shorter entries on sovereignty (主权), registry (登记) and filing (备案), open source (开源), deepfakes (深度伪造) and deep synthesis (深度合成), and alignment (对齐) and value alignment (价值观对齐). Security assessment (安全评估) is also examined within the safety chapter to illustrate how similar language can produce fundamentally different regulatory mechanisms. The matrix below summarises where divergence exists and explains the implications for policymakers in China-facing diplomacy.

*Note: The matrix provides a practical comparison based on the prominent official uses of each term examined in the report. It is not intended as an exhaustive account of how each term is used across every context. The full report explains the methodology and source selection.