Analysis

US officials are working to clarify that Washington has not changed its position on Taiwan, despite President Trump’s comments during the Beijing summit, in which many listeners believe signaled that the United States’ commitment to Taiwan was a subject of negotiation. Secretary of State Marco Rubio told Congress there was “no change” in US Taiwan policy. His comments did not resolve the new ambiguity. President Trump’s discussions with Xi Jinping in Beijing, the recently delayed arms package for Taiwan, and the absence of Taiwan from discussions at the Shangri-La Dialogue have created a new perception problem: Beijing, Taipei, and worldwide partners and allies are no longer just monitoring US policy. Rather, they are trying to shape what President Trump may be willing to negotiate in advancement of his trade priorities.

The clearest example of this shift arose during President Trump’s Beijing summit, where Xi reportedly warned Trump that the two countries could clash over Taiwan if the issue is not handled properly. Eschewing nearly 50 years of nuanced, diplomatic language and precedent, President Trump addressed the issue in characteristically personal terms. Asked whether he would move forward with arms sales to Taiwan, he said it “depends on China” and described the issue as a “very good negotiating chip.” This framing suggests that US material support for Taiwan’s self-defense could become contingent on what Beijing offers in return, re-casting the Taiwan problem as a bilateral issue to be managed personally between Trump and Xi, rather than a fundamental commitment to defend Taiwan’s right to determine its own status. 

The administration then strengthened domestic and international perceptions of a substantive policy shift by delaying approval of a $14 billion arms package to Taiwan. Secretary Rubio tried to clarify that the package is “under review,” rather than formally paused, but this wording  only reinforced perceptions that Taiwan’s defense was now a bargaining chip for the White House. The arms sale serves as a test for the CCP’s ability to shape future US support for Taiwan’s self-defense through pressure, inducements, or concessions. In this context, the United States may be de-emphasizing Taiwan-specific support in favor of broader language in support of deterrence.

The danger for US allies and companies is twofold: Washington has potentially abandoned Taiwan (a worst-case scenario for allies like Japan), or that US policymakers have conveyed to Beijing that the US commitment to Taiwan is for sale. Taiwan is paramount to core US interests: the security of advanced semiconductor supply chains, the regional balance of military power, freedom of shipping, US commitment to democracy, and a peaceful resolution of the Beijing-Taipei conflict. Beijing does not need to force a formal change in US policy; it only needs to create the perception that Taiwan’s fate is negotiable and that the island’s security can be folded into a broader bargain with President Trump. 

For US companies, this perception matters long before a crisis begins. If Beijing concludes that US support for Taiwan is conditional, it will have greater incentive to attempt to change the status quo through military pressure, economic coercion, cyber operations, and other forms of intimidation. Even if Beijing stops short of major military action, heightened PRC pressure on Taiwan could require US firms to reassess their access to chips, reevaluate shipping routes, pressure test contingency plans, and assume greater costs while operating in and around the First Island Chain. Strategic ambiguity, a long-held position of the US government, is starting to undermine–rather than assist–Taiwan’s position, and with it a litany of US military, political, and corporate interests in the region.

Analysis

AI models are the center of a new competition between the United States and China. Citing national security concerns, the US government issued an export control directive against Anthropic’s latest model “to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.” The move took Anthropic by surprise and allegedly came after discussions between Amazon’s CEO and White House officials. Secretary Hegseth’s long-standing demands for unconditional access to Anthropic models, which are part of an ongoing feud between DoD and the company, further reinforce the interpretation that the US government’s continued actions against Anthropic may be more personal than about safety concerns. Irrespective of the existence of any such feud, the directive highlights the sensitivity of foreign access to frontier technology and a desire to harness innovation in service of national security ambitions. 

The USG’s decision to suspend unlicensed access to Anthropic’s most recent models is, ostensibly, a consequence of Anthropic not adequately heeding recently updated government safety concerns. Rumors that Chinese actors managed to access Mythos do not help the company’s case. The onus for resolution appears to be on Anthropic. The administration’s demand that Anthropic now restrict access to foreign nationals (including foreign workers at Anthropic) demonstrates that AI models are now strategic assets to national security rather than simply commercial products. To meet the new requirements, Anthropic must now find an internal team made up of exclusively US citizens, and ensure that non-citizens are unable to access the model without government approval. 

While neither the US government nor Anthropic explicitly mentioned China, the subtext is clear: US-China competition is not only driving decisions within the tech industry, but it is also forcing an increasingly tight fusion between the commercial technology sector and the government in both countries. If emerging technologies will define who wins the new industrial revolution, then the stakes are too high for either government to not be involved. Where the case of Anthropic stands out, however, is that being serious about Chinese access to technology would also mean that the USG would spend an equal amount of energy on banning chip sales to China or other protective measures. Instead, the government is targeting AI models, and by extension, Anthropic, while allowing much greater leeway to chip producers. This discrepancy provides fuel for Anthropic’s claim of being uniquely targeted.

The challenge is that leadership in AI, and by extension national security, is not just a question of who has the best models. It is increasingly a comprehensive, integrated system of talent, funding, energy, data capacity, model diffusion, and support for R&D. Technology development, use, and distribution are now integral to the Sino-American cold war. The competition for Anthropic, for example, is not just other frontier labs in the United States; it is also the entire state, tech sector, and supporting economic apparatus of the PRC. Cultivated talent, heavily subsidized energy, secure supply chains, and R&D through the support of world-class universities in the PRC mean that this is an unfair playing field for US companies. 

This leaves a dilemma for the US government: heavily regulate the tech sector and become a state-company hybrid model akin to China, or cede questions of national security to a private sector that is ill-equipped to take larger national security interests into account or to complete with the resources of the second largest economy on the planet. Neither choice is ideal, but the direction of the US government at the moment is increasingly converging with China’s style of government-controlled innovation, whereby tech giants proceed according to strict government oversight and alignment. At the same time, China is taking notes from the United States. The recent introduction of China’s own export controls signals a tightening on outbound capital and investment (particularly in tech), and increased government control over how PRC citizens finance foreign entities. 

The nationalization of frontier labs will likely become a bigger political issue. Current moves range from President Trump’s talks with OpenAI and a potential "Public Wealth Fund" blueprint, to Senator Sanders’ proposals for an American AI Sovereign Wealth Fund to promote government partnerships with industry. In short, the push for greater government-tech fusion is coming from both sides of the political aisle and is therefore unlikely to abate soon. Collectively, tech companies need to understand that they are building not just profit-making tools, but weapons for national security that governments will view accordingly.

Analysis

The Department of Defense updated its list of known Chinese Military Companies (1260H List), which is intended to identify risks of military-civil fusion (MCF). The changes suggest an expanding scope for what counts as national security interests. We previously reported on a short-lived update to the CMC List in February, which DoD quickly scrubbed from the Internet, but not before the world discovered that Chinese companies Alibaba and BYD had been included. These companies are now confirmed additions to the new June list, along with Baidu, WuXi AppTec, and others. Chair of the House Committee on the Chinese Communist Party, John Moolenaar, issued a statement saying “[these companies’] products should be removed from supply chains our country depends on. American companies must stop doing business with these threats to our national security, otherwise they are enabling China’s military ascendance.”

Until now, the CMC List was largely limited to defense and technology companies, but this new wave of additions further blurs the line between commercial technology and national security. The inclusion of commerce agents (Alibaba), EV producers (BYD), robotics/lidar developers (RoboSense Technology Co.), and pharmaceutical companies (WuXi AppTec) provides a clear signal that the scope of national security is no longer a function of traditional defense concerns. The risk of foreign agents collecting data on US citizens or disrupting commercial supply chains now qualifies. The result is that a greater number of traditionally, purely commercial areas now fall under the rubric of national security, and suppliers or partners who were previously assumed to be safe may now garner heightened scrutiny. Moreover, even if those partners are vetted and prove trustworthy, their potential presence on such a list still poses a reputational risk for US partners in the current political climate.

It is important to clarify that the addition of a company to the CMC List does not mean that a given company now presents an immediate regulatory risk. It does mean, however, that there is significant future regulatory risk. This is because a company’s addition to the List often serves as a red flag for the Bureau of Industry and Security (BIS) and other agencies, alerting them that a given Chinese company is a known Military End User, which increases the likelihood of imposing formal restrictions. Many of the newly listed companies have rejected accusations of military cooperation, but China’s law and practice, including the 2015 National Security Law and the 2017 National Intelligence Law, require that all Chinese companies and individuals provide all requested assistance to Chinese national security institutions. These firms have always presented a significant risk of intellectual property theft or undue influence to US trading partners. The change now is that doing business with these firms increases regulatory and reputational risks to US partners. US companies should use extreme caution when engaging with listed companies, as well as look for decoupling strategies for existing relationships.

The US is not the only country that uses potential MCF risk to manage trade relationships: China has been using the same playbook. Starting from January, Chinese exports of rare earth metals to Japan have dropped by a precipitous 80% on Chinese accusations of the metals’ use in Japanese dual-use products. While the claim is likely politically motivated, as Japanese Prime Minister Sanae Takaichi voiced strong support for Taiwan’s autonomy last winter, the motive is irrelevant to the outcome: If China wants to threaten US companies and their allies, it can do so with dual-use accusations. The takeaway here is that, regardless of such accusations’ veracity, claims of a company’s MCF or production of dual-use products is a tool with ever-expanding applicability and has the potential to upend business as usual.

Analysis

Tech innovation presents the CCP with a political dilemma. Promoting AI is essential to the party’s vision for economic growth, political security, and global superpower status. At the same time, the race to adopt AI and other frontier technologies, promoted in no small part by the CCP’s own demands for China’s private sector, is driving intense competition that is prioritizing productivity over all else. Increased productivity in turn can, at least in the short run, reduce the need for employees in certain sectors.

For Chinese companies already facing involution and declining profits amid heightened competition at home, replacing workers with AI is one solution that seemingly aligns with government directives to embrace chat bots and other technologies. (Involution is a common economic concept in the PRC. It means generating less output over ever increasing inputs.)  Involution is causing a potential nightmare scenario for the CCP. Historically, having a large unemployed population, particularly among young, urban people, spells political disaster for governments. The CCP is well aware that mass AI adoption could mean increased unemployment, and therefore increased social instability. 

In response, the party has made it illegal to fire employees and replace them with AI. AI is intended to boost productivity among existing employees, rather than to remove them from the workforce. This directive belies a nervousness about new technologies that are at once vital for innovation and competitiveness, while simultaneously threatening a society that has already embraced “lying flat” (a socioeconomic rejection of the job hustle) as a response to demands for increased output for less pay or reward. In response, some companies are attempting to avert restrictions by freezing hiring, sidelining contractors, reducing hours, or relying on other measures that effectively minimize spending on employees without technically firing them. Unlike the openness that US tech companies have shown in announcing increased profitability via mass layoffs, the approach in China is more subtle, but nonetheless just as calculated. 

This tension will likely make for an increasingly messy set of signals from China about AI innovation and adoption. Driving new models and diffusing them across the economy will remain a top priority, but PRC tech companies will likely face greater scrutiny about how their models are used and their hiring and firing practices across a range of teams. For global businesses, this will make it challenging to determine China’s ultimate AI goals, and therefore how to best respond.

Analysis

The Five Eyes intelligence alliance warned that Chinese military intelligence services and undercover operatives use platforms such as LinkedIn, Indeed, and Upwork to target individuals with sensitive information. According to the warning, PRC-linked actors allegedly pose as recruiters, human resources consultants, or think tank representatives. They use legitimate-looking job opportunities to identify people with valuable access, build trust, and move conversations from basic research toward the disclosure of sensitive information.

While China also uses traditional intelligence techniques, it scaled its operations from targeted talent scouting to platform wide, automated sweeps on online job sites. By reducing reliance on conferences, business meetings, academic exchanges, and personal introductions, intelligence-linked actors can identify and directly approach people with useful access across government, defense, academia, journalism, think tanks, and industry. Such interactions are also more difficult for employers to monitor, as job sites rely on individual accounts that intentionally separate users from corporate oversight.

What makes this warning easy to overlook is the assumption that only secrets attract attention. In reality, foreign intelligence services often build their collection piece by piece, initially drawing value from data that is not sensitive and restricted, and then moving to increasingly limited and valuable information. People with indirect access to government or commercial intentions, plans, intellectual property, and other intangible assets are of immense intense value. For businesses, even seemingly routine knowledge about contracts, customers, technology, or internal priorities can become valuable when collected, combined, and analyzed at scale.

Companies beyond the defense sector should, therefore, treat suspicious outreach as a diligence issue. Consulting requests, expert-network outreach, and recruiting opportunities may appear legitimate while creating opportunities to collect sensitive company information. Businesses should train employees to recognize and report outreach, review outside consulting and expert-network policies, and assess whether ordinary professional relationships create information risk.