Analysis

The CCP is betting big on AI to solve many of its problems, ranging from ensuring first-mover advantage in new technologies to increasing economic productivity and monitoring dissent. That does not mean, however, that AI rollout is either going smoothly or doing so in a coordinated measure. The recent back-and-forth on OpenClaw, an Austrian-invented autonomous AI agent that has become favored in China, encapsulates tensions at the heart of China’s uneven embrace of AI. Embracing the frontier technology of AI agents unlocks potentially rapid increases in productivity, notably in the push to create “one-person companies” (一人公司). At the same time, embracing AI agents introduces potentially catastrophic security risks from sharing sensitive information with foreign-owned models, like OpenClaw. 

The CCP stressed its aspirations for technology in the 15th Five Year Plan and “AI Plus” (人工智能+) action plan, which emphasized that AI agents should have “deep integration” (深度融合) into 70 percent of the Chinese economy by 2027, 90 percent by 2030, and ubiquitous deployment by 2035. Multiple local government branches took this as a signal that promoting AI adoption should not only be encouraged, but may in fact become a new metric by which the center measures success (and therefore grants promotions and funds projects) among lower ranking officials. The pressure for local authorities, then, is how to rapidly and cheaply roll out AI across local government, businesses, and personal use applications in such a way that the central government can both measure and reward. Seeing the incentive in demonstrating widespread use of AI in their localities, several local governments therefore encouraged OpenClaw uptake, including by offering cash to new users, in part with the hope of demonstrating their prowess in enacting central government demands. Tencent even integrated OpenClaw with WeChat in anticipation of demand.

Few local officials, however, seem to have considered the potential security implications of using a foreign-made AI agent to conduct sensitive government or corporate work, not to mention the standard risks of uploading personal information to a platform with potential vulnerabilities. A quick response by the central government sought to stem the deluge of OpenClaw uptake, banning it from use in state-owned enterprises or official capacities. The challenge is that OpenClaw, or “lobster” (龙虾) as it is affectionately known, is wildly popular. Putting the genie back in the bottle will be challenging for the CCP now that individuals have seen the work an AI agent can produce. 

Despite incentivizing technological innovation, political control is and will remain the CCP’s top priority. The party had loosened its political grip somewhat in the face of anemic economic performance, but that easing appears to be over. Now, the CCP is attempting a challenging balance. The party demonstrated this recently when it banned Manus executives from traveling after they moved the company to Singapore and sold it to Meta. While the party, at least in official channels, initially celebrated Manus’s deal with Meta, the reality is that both developments present intolerable challenges to the party’s control over China’s tech sector. T

What is clear from the grassroots embrace of “lobster” in China is that AI agents will likely become a key part of daily workflows across the world, with Chinese users championing their use. Grassroots enthusiasm for AI agents and agentic AI will spur Chinese tech giants to accelerate rollouts of new platforms and seize first-mover advantage in a fast-paced market. US businesses need to adapt fast to keep pace.

Analysis

The CCP is engaged in a systematic campaign to acquire, replicate, and replace US companies through economic espionage and state-directed industrial strategies. What most US companies don’t realize, however, is that it is not just corporate giants, but also small businesses that are targets. The difference is that small businesses are more vulnerable and less-capable of repelling such attacks. Speaking at the House Committee on Small Business, Tom Lyons, 2430 Group co-founder, testified that small businesses are uniquely vulnerable and ill-prepared for the increasing number and severity of nation-state threats from China. 

Main Street needs help detecting, deterring, and defending against such attacks. But the government also needs to recognize that most small businesses are not competing against other small businesses in China, but against the entire array of resources of the People’s Republic of China (PRC). Chinese firms have major incentives and receive significant assistance from the CCP to steal IP from US counterparts. The potential economic gains from obtaining competitor information by any means necessary produces huge gains at very little cost and risk. Malicious actors also face remarkably few disincentives from the US government, including minimal regulatory and legal consequences for waging economic warfare. 

So what can the US government do to help small businesses? There are multiple, immediate steps that could be effective right away, ranging from criminalizing the receipt of compensation or material benefit from designated espionage organizations, requiring companies to disclose material support or equity investment from countries of concern, or establishing a dedicated counter-espionage program within the Small Business Administration. Other potential measures include cybersecurity tax credits and providing an IP legal defense fund for small businesses. The impediments to such efforts are a lack of understanding among political decision makers and US small businesses alike as to the threat that Main Street faces.  

As we argued in front of Congress: “If this is a war, then Main Street should not have to fight it alone. With the right framework, it will not have to.”

Analysis

On Thursday last week, a bipartisan group of representatives introduced a bill to Congress that would expand and formalize a US ban on sales of advanced semiconductor manufacturing equipment to China and other adversaries, as well as close existing loopholes concerning third party resale and end-user declarations. This is not the first bill proposed to limit China’s access to such technology, but is among the most restrictive. Following the discovery that tech giant SuperMicro smuggled more than $2.5B worth of Nvidia chips and servers to China since US export controls were put in place, lawmakers are finding a renewed urgency to curtail China’s access to US advanced tech. 

While the currently proposed legislation targets only advanced semiconductor technologies, the ever widening scope of such bills should lead companies to plan for how to respond if and when their products are targeted. Whereas the US government previously intervened in companies like Nvidia’s and AMD’s sales of specific products to China, the new bill would become a blanket ban on all related technologies. Tomorrow, the Congress can impose bans on EVs and batteries, navigation tech, manufacturing processes, and a host of other technologies. It could even close certain shipping lanes. In a world where government intervenes in markets quickly and prioritizes national security over industry input, companies need to be prepared to know what their China liability is and what it could mean to lose access to one of the largest markets in the world. 

Even if trade with China is not hindered by the US government, there are also growing signals that the Chinese government will force US companies out anyway. The most recent Chinese government work report makes it abundantly clear that Beijing is aggressively pursuing an agenda of self-reliance that looks to make good on its 2022 directive to “delete America” from its supply chains. In light of the report–and Beijing’s recently opened trade-limiting counter-probes into US practices in China–Chinese Premier Li Qiang’s recent remarks that China welcomes foreign companies and the opening of new markets ring somewhat hollow. Made following the rescheduling of President Trump’s trip to China for May, Li’s statement reads more like a negotiation carrot that can be walked back than a policy change.

All these points show a trajectory that ends in some form of US-China decoupling. While this outcome is not inevitable, it is the path we are currently on. Between the US government’s aggressive restriction of US companies’ sales to China, US lawmakers’ serial proposals to codify in law expansive export bans to China, and Beijing’s insistence on becoming self-reliant with regard to advanced technology and eliminating its US supply chain dependence, decoupling is the logical outcome. 

This does not portend a total and clean break between US and Chinese markets. That is still highly unlikely given the current interconnectedness of markets, but the US regulatory framework for doing business with China is going to become increasingly complex as the government tries to balance national security interests with companies’ access to markets. It is therefore prudent for US companies to map their China exposures in their supply chains and keep abreast of executive and legislative efforts to intervene in US-China trade.

Analysis

The AI conference NeurIPS (Conference on Neural Information Processing Systems) became a new flashpoint for US-China AI competition after conference organizers announced that they would ban paper submissions from affiliates at US government-sanctioned organizations. This would include PRC tech giants, like Huawei. In response, the China Computer Federation (CCF, 中国计算机学会) issued a call for researchers to boycott the conference, and the China Association for Science and Technology (CAST, 中国科学技术协会), a CCP and PRC-government scientific organization, announced that it would suspend funding applications for PRC nationals hoping to attend the conference. CAST also announced that papers included in this year's NeurIPS will not be recognized in CAST applications. Backlash to the conference from both the Chinese government and PRC nationals led to the conference facing intense pressure to reverse the ban, which it did a week after the initial announcement. 

NeurIPS is a California-based organization and is compelled to comply with state and US federal law, including enforcing early bans on submissions by researchers on the US Treasury's ​Specially Designated Nationals List. The organization is caught in a bind: comply with US law and political pressure to restrict access to AI innovation by PRC competitors and government-aligned entities, while maintaining academic openness necessary for free scientific inquiry. Balancing these will likely become increasingly challenging for academic and industry conferences as they face competing demands from the US and Chinese governments. 

These challenges also point to important trends in the confluence of geopolitics with tech innovation. Conferences will likely be increasingly unable to steer clear of political demands for tech “decoupling.” Calls to boycott NeurIPS will also likely make attendance a political test for PRC nationals, even if they are neither affiliated with a sanctioned organization nor based in the PRC. PRC nationals based in the US and working at US tech companies or universities may face political repercussions for attending the conference, including restrictions on access to PRC scientific funding, research collaborations, or personal repercussions against them or their PRC-based family members. The incentive is therefore for PRC nationals to forgo the conference entirely. The challenge for US tech companies is that these restrictions will limit attendance by employees holding PRC passports or collaborating with PRC nationals as a matter of employee safety. 

While this is currently limited to NeurIPS, there is a further risk that such restrictions are extended to all US-based conferences, which has already occurred in other fields. As the CCP becomes increasingly attentive to PRC tech talent, including encouraging PRC nationals working in the United States to “return” to the PRC, conference attendance may bifurcate into PRC-based and US-based conferences with limited participation or interaction between the two.

Analysis

The fundamental challenge in combating economic espionage is that IP theft remains highly lucrative—often so profitable that individuals are willing to risk their careers, their freedom, and their residency to engage in it. Traditional penalties, such as fines or prison sentences, frequently fail to outweigh the massive financial rewards offered by state-aligned organizations or foreign competitors. The US government is now attempting to alter this incentive structure by including revocation of citizenship as a punishment. This week’s court-ordered denaturalization of Yu Zhou and Li Chen, a California couple who stole pediatric medical secrets, is a high-profile example.

By stripping the citizenship of naturalized immigrants convicted of trade secret theft, the DOJ is targeting the one asset that many high-skilled insiders value most. Zhou and Chen, formerly researchers at Nationwide Children’s Hospital, arrived on H-1B visas and spent nearly a decade obtaining citizenship before pleading guilty to conspiracy and wire fraud. The court’s decision to revoke Zhou and Chen’s citizenship on the grounds of "moral turpitude" signals a more aggressive, visible prioritization of denaturalization. 

It remains unclear, however, whether this shift will meaningfully deter the most committed actors. While the threat of losing one's legal status seems as if it would act as a powerful deterrent for those seeking to build permanent lives in the US, so would arrest, conviction, fines, and imprisonment. Yet, loss of citizenship may have little effect on individuals already planning a permanent exit or operating under the direct coercion or protection of a foreign state.  The immediate financial or political rewards of successful theft still dwarf the long-term risk of apprehension and denaturalization.

US companies and research institutions need to understand that some employees are willing to risk everything and to fact that recognition in internal defense strategies. The case of Zhou and Chen reinforces that the most significant threats often come from highly skilled insiders. Because the incentives for theft are so high, businesses cannot rely on federal prosecution to act as a primary deterrent. Internal compliance and security frameworks must be effective enough to detect pattern-of-life shifts and abnormal data access in real-time. If employees believe the rewards for theft justifies the risk of losing their citizenship, a company's internal controls are often the first, final, and only line of defense.

The move toward denaturalization highlights the fact that the US government increasingly considers private sector employees’ citizenship and immigration status to be national security matters.