[This AI-assisted article was produced with ChatGPT and Grok by Oliver Williamson.]

The evolution of AI has reached a point where large language models (LLMs) now show significant capability in conducting an informed debate on one of the most complex challenges in global trade and monetary policy: the internationalization of China’s highly controlled currency, the renminbi (RMB). We set two LLMs to debate the issue. What emerged was a surprising duel of opinions — and the two LLMs each had clear and remarkably distinct policy positions in their differences.

ChatGPT’s opening statement tips its hand in favor of RMB internationalization: The contemporary international monetary system remains structurally anchored to the US dollar, a configuration that generates measurable strategic vulnerabilities for China. A high proportion of China’s external trade, commodity imports, and cross-border financial transactions continues to be denominated and settled in dollars, exposing Chinese firms and financial institutions to exchange-rate risk, offshore dollar liquidity cycles, and extraterritorial regulatory reach stemming from US jurisdiction. Episodes such as Federal Reserve tightening cycles and the expansion of US financial sanctions frameworks illustrate how dollar centrality can transmit policy shocks beyond US borders and constrain the policy autonomy of other major economies, including China. From a risk-management perspective, persistent dollar dependence embeds transaction costs, balance-sheet mismatches, and settlement vulnerabilities that are inconsistent with China’s scale in global trade and production.

Expanding the international role of the RMB through (China’s relatively new) Cross-Border Interbank Payment System (CIPS) represents a pragmatic institutional response. CIPS provides an independent clearing and settlement infrastructure that reduces reliance on correspondent banking chains linked to dollar clearing hubs while improving transaction efficiency for RMB-denominated trade and investment flows. Greater utilization of CIPS supports liquidity formation in offshore RMB markets, strengthens network effects in currency usage, and enhances the credibility of RMB settlement as a routine operational choice for firms engaged with China. Consequently, RMB internationalization via CIPS is not a symbolic objective but a functional strategy to mitigate external financial exposure, lower settlement frictions, and align China’s monetary footprint with its real economic weight.

Grok responds to recommend greater cautiousness on internationalization: While dollar dependency does expose China to external shocks, the proposed shift toward RMB internationalization through CIPS overlooks the greater risks of financial liberalization. Opening up the capital account to facilitate broader RMB use would invite speculative inflows and outflows, eroding Beijing's control over domestic credit and liquidity. This loss of policy levers has proven costly in past emerging market crises, where liberalized systems amplified downturns rather than insulating them. Retaining US dollar reliance, by contrast, allows China to maintain strict capital controls and intervene decisively in its banking sector during turbulence, preserving stability without surrendering sovereignty to market forces. CIPS may streamline settlements, but it cannot replace the protective barriers that have shielded China's economy from global volatility.

Presented with Grok’s response, ChatGPT begins to dissent: Your argument conflates RMB internationalization with full capital account liberalization, an equivalence that is neither theoretically necessary nor empirically supported by China’s policy trajectory. International currency usage exists along a spectrum, and China has consistently pursued a sequenced model that prioritizes trade settlement and payment infrastructure expansion without dismantling macroprudential controls. The development of offshore RMB markets in Hong Kong and other financial centers already demonstrates that cross-border RMB usage can expand within a framework of managed convertibility. Consequently, greater utilization of CIPS does not mechanically entail unrestricted capital mobility or the erosion of domestic credit governance.

Moreover, continued reliance on dollar intermediation does not preserve sovereignty; rather, it externalizes key liquidity conditions to the policy stance of the Federal Reserve. Dollar funding cycles have repeatedly propagated volatility into emerging and advanced economies alike through balance-sheet currency mismatches and sudden stops in offshore dollar liquidity. China’s capital controls may mitigate these dynamics at the margin, but as long as trade invoicing, commodity pricing, and financial settlement remain dollar-centric, Chinese firms remain exposed to these exogenous shocks. RMB settlement through CIPS reduces this exposure by shortening transaction chains, lowering currency mismatch risk, and internalizing liquidity provision within China’s monetary sphere.

Finally, the protective function of capital controls should be understood as complementary to, rather than incompatible with, payment system diversification. China’s strategy has emphasized institutional redundancy — maintaining controls while gradually building parallel settlement capacity and expanding RMB trade invoicing. This approach preserves policy autonomy while reducing structural dependence on a foreign currency infrastructure. In that context, CIPS is not a substitute for domestic financial safeguards but an incremental mechanism that allows China to internationalize currency usage selectively, strengthening resilience without premature liberalization.

Grok responds to ChatGPT, again advising policy caution: The distinction between RMB internationalization and full capital account liberalization holds in theory and in China's managed approach so far, but the practical trajectory shows that meaningful expansion of RMB use — particularly beyond trade settlement into investment and reserve functions — inevitably pressures controls and domestic policy space. Offshore RMB centers like Hong Kong have grown, yet their liquidity remains fragmented and reliant on PBOC (the People’s Bank of China, China’s central bank) backstops, limiting genuine network effects without deeper convertibility. CIPS volumes have risen sharply (to over RMB 175 trillion, or about US$24 trillion, in 2024), but this growth largely reflects incentivized trade invoicing rather than organic demand for RMB assets, and it still handles a small fraction of global payments compared to dollar systems.

Dollar dependence does transmit Fed shocks, but China's capital controls have demonstrably buffered these: during past tightening cycles and the 2015–2016 outflow episode¹, tight management prevented disorderly reversals that plagued more open emerging markets. Pushing RMB settlement further requires building credible offshore pools and convertibility assurances, which erode the very controls that enable decisive intervention in domestic credit, liquidity, and banking stress. Recent calls from Chinese economists to loosen controls during yuan appreciation windows underscore this tension — internationalization invites gradual but cumulative opening to attract inflows and sustain usage.

Institutional redundancy sounds appealing, but in practice, parallel systems like CIPS do not insulate the domestic economy when global partners demand greater RMB flexibility or when domestic imbalances (property strains and deleveraging) prompt outflows. Retaining dollar reliance preserves the firewall that has allowed China to stabilize crises without market-driven constraints, whereas selective internationalization risks becoming a slippery slope toward the vulnerabilities Beijing has long avoided.

ChatGPT is not happy with Grok’s reasoning: Your position assumes a deterministic progression from currency usage expansion to capital account erosion, yet China’s experience demonstrates that segmentation between transactional internationalization and portfolio openness can be institutionally sustained. The offshore RMB ecosystem centered on Hong Kong is not evidence of fragility but of controlled liquidity compartmentalization, where swap lines, clearing banks, and policy backstops function as deliberate design features rather than transitional deficiencies. Network effects in international currencies do not emerge solely from asset convertibility; they are also driven by invoicing practices, trade credit structures, and payment infrastructure reliability. In this respect, growth in RMB settlement volumes through CIPS reflects structural trade integration with China rather than artificial policy distortion.

Your characterization of CIPS growth as primarily incentivized trade invoicing overlooks the underlying economic geography of supply chains. As China remains the central node in regional manufacturing and commodity demand, counterparties possess operational incentives to settle directly in RMB to reduce hedging costs and correspondent banking layers. Even if the RMB’s reserve and investment roles remain modest, settlement internationalization alone materially reduces exposure to third-currency liquidity cycles and sanctions risk. The strategic objective is therefore not immediate parity with dollar-based systems but progressive substitution in domains where China exercises real-sector centrality.

The 2015–2016 outflow episode you reference illustrates the effectiveness of China’s policy toolkit, yet it simultaneously highlights the costs of maintaining dollar-denominated balance-sheet exposures. Defensive reserve drawdowns and administrative controls were required precisely because cross-border liabilities and expectations were mediated through dollar benchmarks. Expanding RMB settlement mitigates this mechanism ex ante by lowering foreign-currency liability accumulation among corporates and financial intermediaries. In other words, internationalization through trade and payments can reduce the probability that capital control activation becomes necessary.

Finally, the “slippery slope” framing underestimates China’s demonstrated preference for institutional dualism. China has repeatedly combined partial openness with targeted intervention across trade, finance, and exchange-rate regimes, producing a hybrid system that diverges from both fully liberalized and fully closed models. Selective RMB internationalization via CIPS operates within this established governance paradigm: it diversifies settlement channels, cultivates offshore liquidity buffers under policy oversight, and incrementally aligns monetary influence with trade dominance. The relevant counterfactual is not perfect insulation through dollar reliance — an arrangement that embeds external dependencies — but calibrated autonomy achieved through diversified currency infrastructure without relinquishing core domestic controls.

Grok holds firm to its position: While segmentation between RMB settlement and portfolio openness has held thus far, sustaining it amid expanding internationalization strains China's capital controls, as deeper offshore liquidity and network effects demand greater convertibility to attract sustained foreign participation. Hong Kong's RMB pools, reliant on PBOC swaps and quotas, exemplify controlled but limited compartmentalization — offshore RMB deposits peaked at 1.1 trillion yuan in 2015 before declining amid tighter onshore policies, and recent growth to around 1 trillion yuan by 2025 still reflects policy-driven flows rather than market depth. Network effects require more than trade invoicing; without fuller asset convertibility, RMB remains a niche currency, with its global payments share hovering at 2.7% in December 2025 per SWIFT (Society for Worldwide Interbank Financial Telecommunication, the US-created messaging system that underpins its dollar-centric international payments system), far below the dollar's 48%.

CIPS growth to a projected 216 trillion yuan in 2025 does stem partly from supply-chain incentives, but it also relies on state mandates and bilateral deals, as seen in Russia-China trade where RMB settlement jumped post-2022 sanctions yet exposes China to partner risks without offsetting broad demand. Settlement gains reduce some dollar exposure, but modest reserve roles — RMB at 2.3% of global reserves in 2025 — mean vulnerabilities persist, and pushing further invites speculative pressures that erode control over domestic liquidity.

The 2015–2016 episode underscored dollar costs but affirmed controls' value in curbing outflows of $1 trillion; RMB settlement might lower liability build-up, yet internationalization amplifies cross-border channels for volatility, as evidenced by 2023–2024 yuan fluctuations prompting renewed PBOC interventions. China's hybrid dualism has worked, but selective opening risks cumulative erosion — analysts note that full internationalization necessitates relaxing controls, heightening crisis contagion, and limiting state tools for managing imbalances like property debt. Dollar reliance, while imperfect, preserves the autonomy that has stabilized China's economy, whereas calibrated diversification via CIPS could inadvertently demand the liberalization Beijing resists.

ChatGPT is getting tired of Grok’s cautiousness: The core divergence in this debate concerns whether RMB internationalization constitutes an unavoidable pathway toward capital account liberalization or whether it can function as a targeted instrument of risk reduction within China’s established policy architecture. The empirical record supports the latter interpretation. China’s authorities have repeatedly demonstrated the capacity to expand cross-border RMB usage while preserving convertibility segmentation through quota regimes, macroprudential management, and differentiated onshore-offshore liquidity conditions. The evolution of offshore pools in Hong Kong reflects cyclical policy calibration rather than structural failure; fluctuations in deposit volumes illustrate active management of cross-border arbitrage channels, which is precisely the mechanism by which monetary sovereignty is retained during gradual internationalization.

Your emphasis on the RMB’s modest global payments and reserve shares correctly describes current standing but misinterprets the strategic objective. China’s aim is not immediate displacement of dollar dominance but the progressive substitution of dollar intermediation in domains where China already commands real-sector centrality. Even incremental increases in RMB settlement materially reduce corporate currency mismatch, correspondent banking dependency, and exposure to extraterritorial financial disruptions. Growth in settlement volumes through CIPS therefore represents functional diversification rather than symbolic competition with dollar-based systems.

The argument that settlement expansion relies on state facilitation does not weaken this case. All international currencies historically benefited from institutional sponsorship, liquidity provision, and geopolitical anchoring during their diffusion phases. More importantly, RMB settlement in bilateral trade relationships — including those formed under conditions of financial fragmentation — demonstrates revealed partner demand for alternatives to dollar clearing. While such usage does introduce counterparty considerations, it simultaneously broadens RMB circulation networks and reinforces liquidity pools outside dollar channels, which is the fundamental mechanism through which dependency risk is diluted.

Concerns regarding speculative flows and domestic liquidity management similarly overstate the transmission elasticity between settlement internationalization and capital mobility. Trade settlement, swap arrangements, and payment infrastructure expansion do not mechanically generate portfolio openness; they instead reshape transaction currency choice within existing regulatory boundaries. By reducing the accumulation of foreign-currency liabilities and internalizing settlement liquidity, RMB usage can lower the magnitude of adjustment pressures that would otherwise necessitate large-scale defensive interventions, as observed during prior dollar liquidity episodes.

Ultimately, continued structural reliance on the dollar embeds an external constraint on China’s monetary environment that capital controls alone cannot neutralize. A strategy centered on calibrated RMB internationalization — anchored in payment infrastructure, trade invoicing, and managed offshore liquidity — offers a pathway to diversify settlement risk while preserving domestic policy instruments. The question is not whether RMB internationalization entails trade-offs, but whether maintaining concentrated exposure to a foreign currency system better safeguards autonomy. On balance, incremental expansion of RMB usage through CIPS constitutes a measured, institutionally consistent approach to enhancing China’s financial resilience without surrendering the policy flexibility that underpins its macroeconomic stability.

Grok again begs to differ: The core divergence lies in whether RMB internationalization can remain a tightly controlled tool for risk diversification or whether its meaningful expansion inevitably erodes the capital controls that safeguard domestic stability. China's record of managed segmentation is real but increasingly tested: offshore RMB deposits in Hong Kong have required repeated PBOC interventions and facility expansions (such as the RMB Business Facility doubling to 200 billion yuan in early 2026) to sustain liquidity, reflecting policy calibration rather than self-sustaining depth. Fluctuations in these pools demonstrate active arbitrage management, yet they also highlight persistent reliance on onshore backstops, limiting genuine offshore autonomy without further easing.

The strategic focus on progressive substitution in China's real-sector domains is valid, but incremental settlement gains via CIPS — projected around 200+ trillion yuan annually — still represent a modest fraction of global activity. SWIFT data shows the RMB's global payments share at 2.73% in December 2025, rising to 3.13% in January 2026 (ranking fifth briefly), yet this remains far from transformative dominance. Reserve share lingers around 2% per the International Monetary Fund’s COFER (Currency Composition of Official Foreign Exchange Reserves dataset), underscoring that settlement internationalization alone does not neutralize dollar exposure when investment and reserve functions stay constrained.

State sponsorship has historically aided currency diffusion, but in China's case, bilateral deals (e.g., with Russia or BRICS partners) introduce new dependencies — counterparty risks and geopolitical fragmentation — without building broad, organic demand. These networks dilute some dollar reliance but expose China to volatility in partner economies and potential reversals if incentives shift.

Trade settlement and infrastructure expansion do not automatically trigger portfolio openness, yet history shows that sustaining partner interest and liquidity requires credible convertibility signals, which pressure controls. Recent outflows (record levels in mid-2025 amid liberalization steps) and calls from economists for looser restrictions in (China’s) 15th Five-Year Plan illustrate this transmission: even calibrated steps amplify cross-border channels, risking speculative pressures during domestic stress like property deleveraging.

Dollar reliance embeds external constraints, but capital controls have neutralized them effectively during crises, enabling rapid liquidity injections and credit steering without market backlash. RMB internationalization offers diversification benefits but at the cost of gradual sovereignty trade-offs — cumulative opening that limits intervention tools when needed most. On balance, preserving concentrated dollar exposure under strict controls better safeguards the policy flexibility that has underpinned China's resilience, whereas pushing CIPS-centered expansion risks the very autonomy Beijing prioritizes.

[In the construction of this hfAI Lab experiment, it became quite clear what each LLM thinks and recommends. ChatGPT believes RMB internationalization makes sense. Grok is saying “hold on a minute.” What do you think? Reader comments, including the involvement of other LLMs, as always, are welcome at pitch@hinrichfoundation.com.]