Not One Bloc
China and Russia are opposite models of state-AI relations, not a shared bloc. The "axis of AI autocracies" framing hides more than it explains.
Start with the lazy consensus, stated fairly. China and Russia get filed under one heading, "authoritarian AI," and the heading is not absurd. Both states want control. Both invoke sovereignty. The problem is that the moment you look at the structures, the single label falls apart.
This issue reads the two structures side by side, in their own languages.
Parallel Frontier is a biweekly intelligence report on China's and Russia's AI ecosystems, read in the original. Each issue works a single primary source in the language it was written in, for people who follow these systems professionally and need the source, not a recap. The name nods to 平行智能 (píngxíng zhìnéng, "parallel intelligence"), a real term in Chinese AI research.
This is the quarterly comparative state-of-play, the two state-AI architectures read side by side, as of Q2 2026, Russia's ownership vertical against China's regulatory mesh. It is a structural comparison, not an events digest.
Russia is a vertical. A handful of labs, Sber as the state-owned champion, a separate Rostec defence track, and "values sovereignty" (ценностный суверенитет) issued from the top, in Putin's November 2025 AI Journey keynote. Control is not laid over the system. Control is the org chart, who owns whom, who answers to whom.
China is a mesh. The field is wide and crowded. Alibaba's Qwen, DeepSeek, Zhipu, Moonshot, ByteDance, Baidu, MiniMax, and a long competitive tail all push hard on capability. The control sits over that field, not inside the ownership of it. It runs through CAC algorithm filings, security review, content and values rules, and party committees embedded in firms. The grid is the instrument. The competition underneath it is real.
That is the grammar of the difference. Russia controls through who-reports-to-whom. China barely controls through ownership at all. It controls through regulation and party embedding across a market that still competes ferociously. Same authoritarian goal, two different machines. Both projects reach for "values." Putin's ценностный суверенитет and China's 价值观对齐 ("values alignment") land on the same word from opposite directions, and that convergence is itself the finding. The divergence underneath it is taken apart in The Translation Gap below.
The punchline is causal, not cosmetic. Comparable sovereignty ambitions, opposite structures, divergent outcomes. China's competition-under-control produces frontier open-weight models, a large share of global open-model usage, and emerging hardware independence. Russia's concentration produces lagging national champions. And the dependency arrow points one way. Russian labs lean on Chinese models and chips. Beijing leans on no one in Moscow.
The boxes are not equivalent, and the comparison has to say so out loud. Sber is a state-controlled bank. DeepSeek is a quant-fund spinout. Alibaba's cloud arm belongs to a listed conglomerate. The chart sets that asymmetry up. The prose has to land it. Making the asymmetry legible is the more honest analysis. Flattening it into symmetric boxes would be the easy mistake.
For policy shops: "axis of AI autocracies" is a filing error, and assessments that treat the two as one bloc will mispredict both. The vertical and the mesh fail in different ways. For labs and ministries watching the export race: the dependency arrow, not the rhetoric, is the tell. Indicators to watch: (1) a national-scale GigaChat deployment abroad, which would validate Moscow's version of sovereign AI (Key Judgment Five), (2) continued Chinese open-model uptake across Southeast Asia, which already validates Beijing's, and (3) any reciprocal Beijing reliance on Moscow, which would falsify Key Judgment Four.
内外有别, "inside and outside are treated differently," is this publication's standing section on what changes when text crosses a language line. Issue 1 caught a state translating itself outward. This issue catches the reverse, what English translation does to two different originals, and the "one bloc" error is manufactured exactly there.
Both capitals reach for "values" as the differentiator. Moscow's term is Putin's ценностный суверенитет, "values sovereignty" (AI Journey 2025). Beijing's is the regulatory term of art 价值观对齐, "values alignment." English coverage renders both projects with the same label, sovereign AI.
The convergence is two authoritarian AI projects independently making "values" the selling point. That much is real, and it is a genuine finding.
What the originals keep. The mechanism, which is where the projects are opposites.
- Суверенитет is a possessive noun. Sovereignty as protection from dependence, the vocabulary of a state that imports capability and fears the dependency.
- 对齐 is a compliance verb. Alignment to mandated standards, enforced through licensing over a market still fighting a price war, the posture of a regulator, not an owner.
- The machinery underneath. Values baked in through ownership and concentration in Moscow. Values mandated through the grid in Beijing.
We assess the shared English label is doing the "axis" work. It merges a protective claim and a regulatory instrument into a single concept and makes opposite machines look like one. Read each term in its own language and the merge comes apart. That is the whole argument of this issue, compressed into two words.
Both terms are elite rhetoric, and rhetoric drifts. The pairing also compresses. 价值观对齐 is one term inside a large regulatory vocabulary, not the whole of it.
Anyone building "authoritarian AI bloc" assessments from English-language sources is working from a vocabulary that has already erased the structural difference. The originals disagree with the label. This publication sides with the originals.
Both capitals reach for "values" as the differentiator, and that convergence is the finding. The divergence in the mechanism underneath it, a possessive noun against a compliance verb, is taken apart in The Translation Gap above. This pairing is the spine of the whole comparison.