AI Competition Between the US and China Escalates
For an extended period, Washington has been captivated by the disparity between American and Chinese artificial intelligence (AI) solutions, as AI has increasingly dominated the landscape of strategic rivalry between these two global superpowers.
The introduction of OpenAI’s ChatGPT in late 2022 signified a significant initial advantage for the United States.
This was bolstered by the initial imposition of export controls on advanced semiconductor chips in October of that year, with authorities estimating that American technology had surged ahead by three years or more.
However, the timeline shrank rapidly as China intensified its developmental pace through algorithmic advancements and an AI strategy characterized by “rapid assimilation.” Notably, this was exemplified by the unveiling of DeepSeek R1 in early 2023.
In recent days, the gap that once seemed formidable has begun to contract, transitioning from months to mere weeks.
This shift is attributed to the emergence of cutting-edge Chinese models from Moonshot AI and Alibaba Group Holding, surprising many in Washington. Notably, Alibaba is the parent company of the South China Morning Post.
Currently, many in Washington characterize US AI policy as being at a pivotal juncture. Three critical inquiries persist: the degree of control to exert over access to leading US models, the necessity of limiting open-weight Chinese models domestically, and the methods to proliferate US AI technologies on a global scale.
“There’s a pervasive uncertainty regarding the situation,” expressed Ryan Fedasiuk, a scholar with the American Enterprise Institute (AEI), who specializes in US-China AI competition.
Impact of China’s AI Achievements on US National Security
China’s recent breakthroughs in AI have surged at a particularly sensitive moment for the AI framework in Washington.
Last month, the Trump administration unexpectedly compelled prominent firm Anthropic to halt access to its latest iterations, namely Fable 5 and Mythos 5, citing national security concerns.
This abrupt maneuver was perceived by some AI experts as detrimental to the international reputation of the US as an AI provider.
This perception was further exacerbated when President Xi Jinping delivered a seminal address at the World AI Conference (WAIC) in Shanghai, where representatives from 28 nations—predominantly from the Global South—endorsed China’s vision of “inclusive” AI solutions.
The implications extend beyond mere leadership in AI technology; they encompass the broader governance inherent in AI systems, as experts caution that advanced AI has the potential to imperil critical infrastructure and vital financial frameworks.
Central to China’s narrative and its challenge to US policies lies its consistent commitment—thus far—to deploying powerful open-weight models that rival their American counterparts.
The perceived threats stemming from advanced US models—triggered by the “unprecedented” cyberattack capabilities linked to Anthropic’s Mythos—prompted the Trump administration’s recent “hands-on” regulatory approach concerning domestic AI firms.
Despite anticipations that China would alter course and curtail its open-source AI strategy once it crossed defined security thresholds, both the Kimi K3, developed by Moonshot, and Alibaba’s Qwen3.8 continue to function as open-weight models, available freely online for a myriad of applications.
This phenomenon has ignited fervent debates regarding whether the United States has relinquished global AI leadership to China by erecting barriers around its own AI frontier, especially following Xi’s endorsement of open-source AI during his WAIC address.
In recent weeks, dissatisfaction has burgeoned among users of leading models from OpenAI and Anthropic, particularly regarding the latest restrictive measures imposed by these firms.
While these companies assert that restrictions are designed to thwart harmful endeavors, such as facilitating cyberattacks, users have encountered limitations that hinder routine tasks, including coding and AI research.
“Kimi K3 has effectively resolved 15 critical security flaws that Codex and Fable neglected due to their ‘cyber guardrails,’” remarked David Sacks, a former White House AI adviser, in a social media commentary. “There’s no rationale for inhibiting American models on tasks that Chinese systems manage effortlessly.”
While previously justifiable for the US to regulate frontier AI as a “rare commodity,” the rise of advanced Chinese open-weight models suggests that this strategy may ultimately prove ineffective, noted AEI’s Fedasiuk.
“A robust US strategy must adapt to a reality where AI is affordable, ubiquitous, and uncontrollable,” he stated.
Currently, proponents assert that the US should shift to a more “proactive” strategy aimed at bolstering AI infrastructure in countries such as the United Arab Emirates and Saudi Arabia to facilitate the global deployment of US AI models.
This transition marks a shift from a traditional competition over advanced methodologies to one focused on comprehensive infrastructure development.
Despite US export controls, which have not halted Chinese advancements, they do inhibit the pace of development by constraining computational resources.
Recent reports indicate that Moonshot was compelled to suspend new user subscriptions shortly after launching its powerful model, due to computational limitations—the latest in a series of Chinese tech firms reporting similar deployment issues.
Additionally, uncertainty looms regarding how Washington will oversee these potent Chinese open-weight models domestically, especially as they have gained traction in the US over the past year.
Many within Washington’s national security apparatus advocate for the imposition of restrictions on Chinese models, asserting that they present considerable risks to the supply chain and broader security landscape.
Conversely, figures like Sacks and former administration officials, including Sriram Krishnan, have argued for a more permissive approach to allow the proliferation of American open-weight AI, offering Kimi K3 as a cautionary tale regarding the consequences of overly restrictive policies.
In recent days, the chasms in US AI policy circles have become increasingly evident, with reports indicating that Kimi K3 has emboldened certain factions within the Trump administration to propose broad-based bans on Chinese open-weight models, while other officials advocating for “pro-competition” policies oppose these measures.
“Implementing some restrictions on Chinese model adoption could be justified given the risks of ideological bias and clandestine agents,” stated CNAS’ Hayum, referencing new research suggesting potential “secret loyalties” within AI models that inadvertently promote the interests of entities regarded as adversarial, such as China.
Trump Administration’s Position on Open Source Models
On Tuesday, Treasury Secretary Scott Bessent asserted that the Trump administration “endorses open source models,” while threatening sanctions against Chinese entities found to be illicitly training AI models utilizing outputs from US systems, a process commonly referred to as “adversarial distillation.”
He also suggested during an appearance on Fox Business that US companies might eventually be obligated to disclose their utilization of Chinese models to their clientele.
Wang Yaqiu, an expert at the Penn Project on the Future of US-China Relations, has posited that Washington may be inclined to underscore the ideological differences separating US and Chinese models, particularly as technical capabilities become increasingly analogous.
Nevertheless, questions persist regarding the Trump administration’s commitment to such an approach.
Previously, the administration had tasked the Centre for AI Standards and Innovation (CAISI), an organization operating under the National Institute of Standards and Technology, with evaluating Chinese models based on their “alignment with Communist Party rhetoric and censorship.” However, the agency has ceased to publish such evaluations in recent reports.
Furthermore, a spokesperson for the US Department of Commerce confirmed that CAISI’s director, Chris Fall, departed from the role after merely three months, raising concerns regarding the future operations of the AI oversight body.
According to CNAS’ Hayum, a pivotal risk confronting US national security policy at this crucial moment in the US-China AI rivalry is persistent indecision: “If the Trump administration opts for ambiguity and fails to establish a clear vision moving forward, it would represent a missed opportunity.”
The White House and the Department of Commerce have yet to respond to inquiries for comment.
Source link: Thestar.com.my.
