[{"data":1,"prerenderedAt":92},["ShallowReactive",2],{"post-ai-truce-progress-or-political-theatre":3},{"id":4,"title":5,"aeo":6,"author":6,"body":7,"date":77,"description":78,"draft":79,"extension":80,"geo":6,"image":6,"keywords":6,"meta":81,"navigation":82,"ogImage":6,"path":83,"seo":84,"seoDescription":85,"seoTitle":5,"sitemap":86,"stem":87,"tags":88,"__hash__":91},"posts\u002Fblog\u002Fai-truce-progress-or-political-theatre.md","The AI Truce: Progress or Political Theatre?",null,{"type":8,"value":9,"toc":68},"minimark",[10,17,20,23,28,31,35,38,41,45,48,51,55,58,62,65],[11,12,13],"p",{},[14,15,16],"strong",{},"INSIGHT  ·  GEOPOLITICS & TECHNOLOGY",[11,18,19],{},"INSIGHT N° 59  ·  04 November 2026",[11,21,22],{},"All eyes were on the United States when President Trump hosted his Chinese counterpart during a state visit in late September. The itinerary featured a White House military review, Oval Office talks, a high-profile state dinner with technology leaders in attendance and a joint tour of foundational American democracy documents at the National Archives. Out of all the events in Xi’s visit, it was the state dinner which caused a wave of interest, but why? The historic state dinner featured a very prominent guest list, including leaders from Nvidia, Microsoft, OpenAI, Google and SpaceX. The guests in attendance sparked the perception that this dinner was more of an AI summit than a diplomatic one. Amid growing concerns about AI safety and its rise to the top of the global geopolitical agenda, this summit was placed under a much closer microscope than expected. The visit wrapped up with limited deliverables, including an extension of the trade truce; however, only for a period of two months, much shorter than expected. However, on the AI front, both countries agreed to hold an AI dialogue within two months and set up a communication channel for AI incidents. While it contains potential, it remains a small step, yet an overbearing question remains: what actually changed in the AI race, and what aspects only appeared to change? The answer to this doesn't lie in the room where the leaders met, but in the places the summit’s agenda doesn't reach.",[24,25,27],"h2",{"id":26},"two-leaders-two-vocabularies","Two leaders, two vocabularies",[11,29,30],{},"Before leaving the room, it is crucial to understand how the two leaders spoke about AI. According to the Centre for American Progress scorecard, President Xi called for keeping AI under human control, while Trump countered by saying that fears of rogue AI were a hoax. Following the summit, Trump announced that major tech companies had signed an accord to self-police AI development through internal and external reviews. This clearly indicates that, in this context, safety is framed as voluntary and written by the industry it covers. While this move isn't dismissive and the fact that voluntary commitments can still shape norms, it prompts further thought about who enforces a review that nobody is required to pass. Moving from safety norms towards cooperation, Trump said he did not want to do anything further on it, because the United States is leading in the race, yet another indicator of uncertainty. When one of the parties championing the corporation explained that he saw little need for it, it immediately set a precedent for future uncertainty. Amidst this spell of ambiguity, in order to know what either side truly believes, there is no choice but to follow the chips.",[24,32,34],{"id":33},"chips-control-regime-vs-reality","Chips: Control regime vs Reality",[11,36,37],{},"Since the Biden administration, the policy of using export controls to block China from acquiring advanced US chips for frontier AI training has been prominent, and it is still maintained under the current Trump administration. Tracking where restricted chips actually move and how they are used provides a clearer picture of the policy’s practical effects. US rules restrict the sale and",[11,39,40],{},"export of advanced Nvidia and equivalent chips to Chinese entities. The administration’s aim is to slow the development of frontier AI systems that rely on that hardware. The specifics of the rule primarily target physical shipment and ownership of the chip, but do these restrictions prevent shipment of restricted chips? Many investigations and enforcement actions have identified numerous networks that route restricted chips through third countries, predominantly in Southeast Asia. Multiple public cases include seizures of cargo and property linked to the trade, as well as indictments involving hundreds of millions of dollars in servers containing Nvidia A100, H100, and other related processors. It is cases like these that demonstrate that diversion occurs; however, they do not establish the total volume of chips that ultimately reach Chinese data centres or training clusters. Currently, regulations govern the physical movement of the chips, but unfortunately, they do not restrict their remote use once installed in data centres outside China. This gives Chinese firms an avenue to rent compute capacity on advanced Nvidia hardware located in third countries such as Thailand, Malaysia or Singapore. Official statements and reporting have cited many instances in which Chinese AI companies have accessed high-end Nvidia systems, including Blackwell-generation chips, through overseas facilities. Under the existing rules, this access remains completely legal, which brings the greater question: how does this impact US suppliers? Nvidia has openly disclosed in SEC filings that export controls have effectively closed it out of China’s data-centre market. The company has further noted that this exclusion has allowed its competitors to expand their developer ecosystems and customer bases in that market. As a result, the US firm incurs commercial costs while compute resources remain available to Chinese users through both physical diversion and remote access. But if China can still access restricted chips and US firms are simultaneously losing market share, are the Chinese accelerating their efforts to develop more substitutes domestically?",[24,42,44],{"id":43},"the-substitution-effect","The Substitution Effect",[11,46,47],{},"When export controls limit access to advanced foreign chips, they can also accelerate Chinese firms' efforts to develop and adopt domestic alternatives. Ultimately, this leads to two outcomes: some restricted hardware still reaches users through diversion or remote access, while parallel investments in the domestic semiconductor sector continue to grow. In April 2026, DeepSeek released and open-sourced its V4 model series, which includes its V4-Pro and V4-Flash variants. One major change was that DeepSeek adapted its AI model to work with Huawei’s Ascend chips. Huawei confirmed that its chips support the model and were used to train part of V4-Flash. This includes advanced computing systems built using Ascend 950 processors. Prior to this change, DeepSeek models heavily relied on Nvidia hardware. Multiple Chinese chip vendors, including Huawei, ensured that DeepSeek’s V4 model could run",[11,49,50],{},"on Chinese-made chips from the very beginning, enabling these chips to handle tasks such as running the AI model and processing requests. Further reports in September 2026 have cited investor discussions and DeepSeek founder Liang Wenfeng, stating that the company’s major strategic priority is to train Huawei and other domestic chips. According to those accounts, Huawei's deliveries of training-oriented chips could begin as of late 2026 or early 2027. However, these statements remain mere reports, so time will tell whether further steps will be taken to improve the domestic stance on chip production. As of now, these steps indicate a progressive shift towards Chinese-designed accelerators and supporting software, while the continued availability of restricted compute persists through existing channels. It is safe to say that the net effect has reduced exclusive dependence on US chip suppliers for certain AI development and deployment workloads, while domestic hardware and software capabilities continue to expand and mature. It is understood that China is finding workarounds, building alternatives and reducing its reliance on US chips. So, can another round of diplomatic dialogue really change the game, or are both sides talking past the problem?",[24,52,54],{"id":53},"why-cant-dialogue-fix-this","Why can't dialogue fix this?",[11,56,57],{},"The newly established bilateral channel and Super Intelligence dialogue function primarily as an incident notification hotline. It is designed purely to flag specific security-related events rather than to negotiate structural questions of chip supply, remote compute access, or the cross-border spread of AI technology. The scope of this mechanism is not the only factor; many misaligned incentives can also deter the solving of core issues. This can be linked to the earlier discussion of the two leaders' differing perceptions. Public statements from US leadership emphasise the inessentiality of cooperation on deeper AI integration, solely because the United States remain ahead in the race. Yet, independent reporting indicates that the performance gap between leading US and Chinese models has narrowed to only a few percentage points across major benchmarks. The coming months will reveal whether dialogue can keep pace with the technology race, or whether the gap between diplomatic promises and technological realities continues to widen.",[24,59,61],{"id":60},"the-calendar-close","The Calendar Close",[11,63,64],{},"Several high-profile dates now sit towards the end of the calendar. Both Presidents Trump and Xi are expected to attend the APEC leaders’ meeting in Shenzhen in November and the G20 summit in Miami in December. Additionally, the first formal session of the new US-China Super",[11,66,67],{},"Intelligence Dialogue is scheduled for November, and, in parallel, reports indicate that Huawei may begin delivering new training-oriented chips by the end of 2026. These meetings and delivery windows will soon form a practical scoreboard, in which the key signal will be whether questions about advanced chip access or compute availability will appear on any formal agendas. Perhaps if they remain absent, the underlying hardware constraints will largely remain unchanged, while diplomatic engagement proceeds on other tracks. To summarise, the leaders met in one room, yet the race is being run in all the others; unless November’s dialogue puts compute on the table, it is safe to assume that this race and its subsequent hardships are continuing.",{"title":69,"searchDepth":70,"depth":70,"links":71},"",2,[72,73,74,75,76],{"id":26,"depth":70,"text":27},{"id":33,"depth":70,"text":34},{"id":43,"depth":70,"text":44},{"id":53,"depth":70,"text":54},{"id":60,"depth":70,"text":61},"2026-11-04","The US-China AI dialogue promises a new channel for managing risk. But export controls, overseas compute access, and China’s domestic chip push reveal how much of the technology race remains beyond the room.",false,"md",{},true,"\u002Fblog\u002Fai-truce-progress-or-political-theatre",{"title":5,"description":78},"An analysis of the US-China AI dialogue, chip export controls, remote compute access, and China’s drive to build domestic AI hardware.","[object Object]","blog\u002Fai-truce-progress-or-political-theatre",[89,90],"Geopolitics","Artificial Intelligence","6WhyOIaM7yZOFCDAO_h7zxGrfPtfuGeDMf18jGWyfl4",1793783155918]