This excellent ChinaTalk interview with Logan Wright reminds me of an @elyratner Foreign Affairs essay (foreignaffairs.com/china/fau…) from a few months ago that, among other things, argued for strengthening chip export controls on China. Denying chips and manufacturing equipment would constrain Chinese capability today while substantially raising the cost and slowing down the chip indigenization program funded by an increasingly strained financial system. It would sharpen the tradeoff facing Beijing between its AI ambitions and other priorities.
Agree, one of the best episodes of any China podcast in quite some time. Raised as many questions as answers, and kudos to @jordanschneider for curating and navigating them.
Daniel Remler retweeted
I wrote about the systemic failure by the world's most valuable company and powerful government to stop smuggling of AI chips to China
US officials say Nvidia is missing obvious signs. Nvidia says "companies honor US laws as they are, not as America’s critics wish they were"
1/
Daniel Remler retweeted
It seems rumors of an US-China AI dialogue and crisis line were greatly exaggerated.
- Beijing hasn’t confirmed it
- The crisis line isn’t a crisis line
- The dialogue may not be a dialogue
There were already low expectations for this summit. This takes them even lower.
Scoop: Treasury Secretary Scott Bessent's proposed AI safety "notification mechanism" is simpler than that moniker suggests: It's an open channel between him and Chinese Vice Premier He Lifeng, four people familiar told me, @PhelimKine and @_AriHawkins. politico.com/news/2026/09/22…
Unintentionally clever—the companies can't respond in the affirmative without crossing Beijing (which, needless to say, they won't do), and responding negatively would pierce the illusion among some in the West that there is a meaningful constituency for AI safety in China. If the Chinese companies have a creative writing major on staff maybe they could thread the needle with a response, or maybe Beijing will issue a non-response, but most likely this will be met by deafening silence.
NEWS: @RepRoKhanna writes the CEOs of Alibaba, DeepSeek and Moonshot asking them to cooperate in an intl effort to pace frontier AI.
+ Khanna will be hosting an emergency hearing on global AI pacing policy next Weds
washingtonpost.com/wp-intell…
The definitive take on the upcoming U.S.-China AI talks by one of the only people in the world who has actually done it—I’m biased as I worked for him and staffed the 2024 talks, but anyone with an interest in AI diplomacy needs to be reading Seth Center
: nytimes.com/2026/09/16/opini…
“Beijing is going into these talks viewing AI safety not as a shared humanitarian mission, but through an adversarial lens” 🎯 exactly right from @henrysgao
Glad to speak to @lilkuo in this nice @nytimes story to unpack the AI risks article by China's Minister of State Security.
nytimes.com/2026/09/14/world…
This is not an outlier stance by a rabidly nationalistic outlet, or even particularly new—it’s the consistent subtext and sometimes text of official PRC AI diplomacy going back to at least 2023.
The @globaltimesnews WeChat commentary on calls for a coordinated slowdown is quite something.
Says the calls for slowing down AI development are reflective of a Cold War containment strategy, and that the “real purpose” is allegedly to use technological barriers and rule-setting monopolies to contain PRC’s AI development, preserve U.S.’s “monopolistic hegemony” in frontier tech, and exclude the PRC from global AI governance.
Daniel Remler retweeted
An essay by Anthropic's CEO calling for a slowing down of AI development may look like a rational statement focused on safety risks, but it is really a ‘Cold War playbook’ targeting China, the state-backed Global Times said in an editorial reut.rs/4xToVsT
Good points from @b_judah. I’d add the Chinese today are practitioners of Kissinger-style linkage in AI safety today as in many other fields and understand that history very well.
However in two crucial way when it comes
to US-China AI safety dynamics the dynamics are utterly, spectacularly, different from the Cold War.
First up: it’s important to establish how different the current US situation is. In the Cold War nuclear arms development and expertise was inside the state, influenced only by the government’s interests and agenda. This time AI expertise and development is taking place outside the state influenced overpoweringly by the prospect of trillion dollar IPOs. In the Cold War those influencing the President on nuclear weapons had no commercial motives to develop more of them.
Secondly, on the differences, the AI community speaks frequently about how “China means we have to keep going.” But there is no security instruction from men in grey suits in the US intelligence community to keep going to these firms. Instead there was a political instruction from the White House to ignore AI safety because a group of commercially interested investors and oligarchs, many of whom have now changed their minds, convinced Trump this was nonsense that would damage the economy and benefit China. This is the fusion of Trump politics and Silicon Valley not the fusion of the Cold War state and the military-industrial complex.
Daniel Remler retweeted
Zeiss advanced optics were one of the most critical chokepoints controlled by the U.S. and allies. This was one of the major reasons China hasn't been able to catch up in EUV yet.
Zeiss lenses were presumed by many to be covered by existing U.S/Dutch controls, since Zeiss is an exclusive supplier to ASML. But it seems that this no longer holds. This is no longer a commercial matter but one critical to national security.
The U.S. should prioritize efforts to coordinate export controls with Germany to keep Zeiss lenses out of China's hands.
Remarkably rich threat intel report from @AnthropicAI. This is just the tip of the iceberg—there is likely a much wider range of activities the PRC and possibly other actors is pursuing using Chinese open weight models that no American company can or will be able to detect.
We're publishing our most detailed threat intelligence report to date.
It covers how people tried to misuse Claude—for cyberattacks, influence operations, surveillance, biology, and building weapons—and how we found and stopped them.
We disrupted every operation in the report, and used the lessons from them to strengthen our safeguards. Where appropriate, we also shared what we found with authorities and other AI companies.
These cases are not typical: we’re highlighting some of the most sophisticated misuse we’ve seen. But they’re especially important to discuss, because they show us where AI misuse is headed, where our safeguards work, and where they need to improve.
We’re publishing this report so others can spot the same activity on their own platforms, and so we can give the public a clearer view of how emerging threats develop.
Read the report: anthropic.com/threat-intelli…
This is not just the best way to slow down China's AI development in the abstract, but to limit the CCP's application of AI for the Chinese military, intelligence services, and surveillance state that directly threaten Americans every day.
This quoted post is unavailable.
It's interesting to me that for some proponents of U.S.-China AI cooperation, like @pstAsiatech below, mistrust is always and only the fault of the United States and the onus is on Washington to fix it. It's always "Blame American AI First," to paraphrase Jeanne Kirkpatrick.
Hard to see how the United States is supposed to engage with China on AI cyber risks when the CIA Deputy Director is saying that an APT operated by the MSS remains a “very critical threat.”
Major cybersecurity advisory on adversarial distillation with extensive and detailed attribution of PRC AI developers. With these facts established, the U.S. government should permit U.S. companies to work together to detect and disrupt these campaigns, as the report recommends.
China-based AI companies are illicitly distilling U.S. frontier AI capabilities. Read NSA’s new report, co-sealed with @FBI, @CISAgov, highlighting AI knowledge distillation, TTPs used, and recommended mitigations: media.defense.gov/2026/Sep/0…
For more recommendations on how to defeat these activities and details on how PRC entities obfuscate their activities using an ecosystem of intermediaries, check out my paper from June with my colleague @BenHayum cnas.org/publications/report…
Daniel Remler retweeted
A Malaysian decision to use Huawei chips for a new sovereign AI data project would implicate Trump's "ART" trade deal with the country. Let's unpack:
1. According to Bloomberg, Malaysia may use Huawei Ascend chips for a $500 million sovereign data initiative managed by a state-linked telecom company. The stated reason for preferring Huawei over US tech is perceived US IT risks, including the "CLOUD Act," which lets US courts to compel tech firms to hand over info to the government.
2. Trump's "ART" Trade Deal with Malaysia includes several provisions relevant to Malaysia's sovereign AI data deal:
Art. 6.1 provides that Malaysia shall "facilitate and promote investment by the [US] in sectors including...telecommunications...and infrastructure services."
Art. 5.2(2) states that "Malaysia shall cooperate with the [US]...with a view to restricting transactions of its nationals with individuals and entities" on the U.S. "Entity List" (which includes Huawei).
Annex Section 5 provides that "Malaysia commits to only using communication technology suppliers that do not compromise the security, safeguards, and intellectual property of ICT infrastructure....[And] shall ensure that [transportation infrastructure] use[s] digital logistics platforms that provide...protection against national security risks...."
3. These provisions do not specifically single out "AI data centers." That said: (A) AI data centers should fall within the positive obligation to enable U.S. investment in the "telecommunications" and "infrastructure sectors," which would imply Malaysia should give a US firm a preference. (B) The obligation on export controls, by its text, does not apply only to cooperating to restricting *exports* to Entity List companies, but to "restricting transactions of its nationals with" Entity List companies. Picking Huawei chips for a sovereign AI data center project is not "restricting transactions" with Huawei. (C) If the AI data center involves communications or logistics data and/or information processing, it would seem to contradict the Annex commitment to use only trusted technology.
4. The ART provisions are effectively unilateral and discretionary, e.g., the Trump Admin can raise a potential violation if it wants to, and if it decides there is a violation, it can impose additional tariffs. (Assuming USTR can find a U.S. domestic law basis for the tariffs). So if Malaysia does go ahead with the Huawei deal, the Trump Administration will have to decide whether it views the deal as an ART violation, and, if so, what to do about it.
5. The deal raises important strategic questions about how valuable the ART "economic security" provisions will be. If the ART helps push Malaysia away from Huawei (even if quietly), it'll be a win for the ART. On the other hand, if Malaysia buys Huawei and Trump just lets it slide, that will raise questions about how serious and durable the ARTs will be.
Important piece from @AnaSwanson et al. on the role of Inspur/Aivres, Maginfra, and other PRC-linked companies in enabling the massive circumvention of U.S. export controls—a stronger BIS affiliates rule + remote access restriction would go a long way to closing this loophole.
New: Washington repeatedly blacklisted this Chinese tech giant. But through a network of global subsidiaries, the company has reinvented itself and kept shipping America's best chips to feed Chinese AI
nytimes.com/2026/09/06/techn…
with @paulmozur @trippmickle @KeithBradsher
Very interesting—@MrKRudd argues the CCP believes AI can resolve the socialist calculation debate and replace the market as the most efficient means of allocating resources in the economy. Doesn’t seem like a party much concerned about “loss of control.”
If you read Xi Jinping’s ideological framework carefully, my three-line summary of the ideological changes he’s brought about in the Chinese system is as follows: mrkrudd.substack.com/p/the-e…
The notion that we should cooperate with China to monitor AI-enabled attacks and asking Chinese companies to “police themselves” to prevent AI-enabled cyberattacks—when China is itself the greatest perpetrator of cyberattacks against the United States—is laughable.
Some new information from Reuters on the AI safety discussions between the U.S. and China. Given the current climate it's seeming more and more likely that some new rules and agreements will emerge from this.
- Both sides have discussed AI guardrails at a Track 1.5 dialogue in Beijing last week
- 'Chinese officials have repeatedly stressed the importance of the AI talks in preparatory meetings with U.S. counterparts, and view them as a major deliverable of the U.S.-China leaders summit'
- 'The U.S. wants to discuss cooperation on monitoring AI-directed cyberattacks, and has floated a proposal to ask U.S. and Chinese AI labs to "police themselves" and share information to prevent AI-linked cyberattacks.'
- 'In the past week, China's cyberspace regulator publicly warned about "extreme AI loss of control risks" and a state media-affiliated blog criticized Anthropic, emphasizing that any limits on frontier AI should apply equally to Chinese and U.S. models.'
- "The Chinese side has expressed concern around whether the U.S. has sufficient regulation around the most advanced AI models. Both sides are motivated to make sure they can manage a cross-border crisis effectively," said Scott Singer, co-director of the China AI Initiative at the Carnegie Endowment for International Peace.
