Chen Yixin Discusses AI Risks, Subtly Acknowledges CCP's Free Ride Exposed by the U.S.

OpenAI diagram (created by ChatGPT)

[People News] In September 2026, Chen Yixin, the Minister of China's Ministry of State Security, published a signed article in the magazine 'China Cybersecurity', where he elaborated on the six major risks associated with artificial intelligence. The primary concern highlighted was the 'systemic impact on the political security environment', followed by issues such as cyber offense and defense, theft and leakage of confidential information, technological monopolies, impacts on social governance, and transformations in military forms, among others.

The article ultimately underscores the importance of maintaining the Party's comprehensive leadership, asserting 'Party control over the internet and data'. It claims that the aim is to convert the Party's leadership advantages into effective AI governance, establish a secure regulatory platform, and enhance efforts on core technologies like computing power chips. Concurrently, Chen Yixin indirectly criticized 'certain countries' for leveraging their technological advantages to impose restrictions, disrupt industrial chains, and worsen the imbalance in the development of the intelligent industry.

Notably, just a few days earlier, U.S. law enforcement and intelligence agencies (NSA, CISA, FBI) jointly released a statement naming six Chinese companies: DeepSeek, Dark Side of the Moon, Alibaba, MiniMax, Leap Star, and Zhipu. They accused these companies of extracting cutting-edge AI model information from the U.S. on a large scale through 'industrial models' using distillation methods to expedite the training of their own products, and indicated that this was 'likely done with the knowledge of the Chinese government'.

Anthropic has released a 154-page threat intelligence report that reveals even more shocking details. These companies have utilized thousands of fake accounts, proxy networks, cross-session replay, and even discreetly forwarded their own user requests to Claude, extracting the reasoning processes of thought chains to train their own models.

Alibaba has been identified as the largest offender, with over 151 million conversations occurring in just a few months. The Dark Side of the Moon forwarded Kimi user requests to Claude without users being aware. Even more absurdly, the content forwarded from Chinese AI includes surveillance footage of PLA-related facilities, internal keys from state-owned enterprises, case management logic from the public security system, and even defense-related credentials from Russia, all while Chinese companies themselves sent sensitive data to American servers.

Anthropic's intelligence report has drawn global attention and created a significant public outcry, exposing the fact that the AI technology the Chinese Communist Party (CCP) boasts about is essentially a case of the CCP taking advantage of American AI. This has been a largely unacknowledged reality that the CCP has consistently denied.

In response, the Chinese Ministry of Commerce stated that distillation is a standard practice in the AI field and should not be politicized, accusing the U.S. of engaging in technological hegemony, monopolizing computing power, and applying double standards. They warned that if this leads to the suppression of Chinese companies, China will take countermeasures. The Ministry of Foreign Affairs also quickly supported this stance, emphasizing that Chinese AI represents the achievement of 'high-level technological self-reliance and self-improvement.'

The recent exposure of the Chinese Communist Party (CCP) came as a surprise and clearly angered the highest authorities. Reports suggest that 16 executives from a prominent AI company were arrested immediately. Following this, Chen Yixin from the Ministry of State Security took the initiative to address the situation. The series of actions taken reflects the CCP's contradictory mindset after a significant setback: they wish to retaliate while also trying to rectify the situation, projecting a tough exterior while actually feeling panicked.

Chen Yixin emphasized 'Party control over AI,' essentially attempting to reinforce political control in the face of the crisis and to counteract negative public sentiment. He remarked, 'Some domestic users lack security awareness, using foreign AI products to handle sensitive information and exporting data abroad, which has led to large-scale data leaks within the country.' This statement indirectly acknowledges that domestic companies' 'free-riding' on American models has spiraled out of control, ultimately jeopardizing China's own security.

This situation represents the latest iteration of the CCP's longstanding approach to technological development: politicized operations, Great Leap Forward-style advancements, and systematic appropriation of Western technology, with blame shifted whenever issues arise.

First, let's examine the politicization aspect. The main risk identified in Chen Yixin's article is not the technology itself, but rather the 'hostile forces' that utilize deepfakes and AI-generated bots to create political rumors and engage in cognitive warfare, thereby threatening political security, institutional security, and ideological security. This narrative is a staple for the CCP; any new technology must primarily serve the security of the regime rather than simply fostering innovation or market growth. AI has been explicitly categorized under Party control, continuing the tradition of Party oversight over media and data.

This indicates that algorithms, models, and data pools must all be subject to political scrutiny; the concepts of open-source and openness are merely slogans, with actual control firmly in the hands of the Party. This prioritization of political considerations leads to distortions in the industry, where companies are pressured to seize opportunities and benchmark against the United States while simultaneously having to operate within the constraints of censorship, leaving virtually no space for genuine original innovation. Consequently, while there is a plethora of large models appearing on the report card, the actual quality is worrisome, as it heavily relies on external distillation and the transfer of computing power to circumvent sanctions.

Examining the Great Leap Forward-style development once more, the Chinese Communist Party's political valuation of AI mirrors its previous expectations for steel, chips, and new energy: a focus on resource concentration, a nationwide system, and achieving rapid advancements in a short timeframe. From the 'Eastern Data, Western Calculation' initiative to the 'independent' breakthroughs in computing power chips, and from the competition in large models to the shift towards intelligent systems, the official narrative consistently emphasizes that 'opportunities and challenges coexist, but we must seize the initiative.'

Nevertheless, the fundamental issue lies in the foundational theories, structural innovations, and high-end computing power, which have long been constrained and dependent on external influences. As the United States tightens its chip export controls, Chinese companies quickly pivot to Southeast Asia for training and intermediary access to American models. Distillation, originally a neutral technology in the industry, involves teacher models training student models, which can lower costs. However, if it becomes overly audacious, transforming into industrial-scale models with covert and unauthorized fake accounts flooding the market, along with user traffic hijacking, it signals that a complete collapse is imminent.

The Anthropic report indicates that these operations are not random occurrences; they are systematically organized and specifically target high-value content such as thought chains, even switching objectives within 24 hours following model updates. This is not a case of interactive learning or borrowing; it is clearly a form of systematic one-way freeloading. The Chinese Ministry of Commerce claims that 'American companies also distill Chinese models,' attempting to draw a false equivalence while sidestepping a crucial distinction: most Chinese open-source models are based on existing public results or distilled outcomes, whereas the core barriers of cutting-edge American models stem from substantial computing power and the accumulation of raw data, along with service agreements that explicitly prohibit competitive training. Framing contract violations as a 'common practice' is a rhetorical maneuver that misrepresents freeloading as an industry standard, revealing a fundamentally rogue character.

The act of plagiarizing Western technology is central to this approach. From high-speed rail to photovoltaics, from 5G to chips, and now to large models, the typical path for technological advancement pursued by the Chinese Communist Party (CCP) involves: importing and digesting technology, coercing intellectual property transfers, acquiring technology through overseas purchases and theft, and violating regulations to manipulate talent flow. Ultimately, they proclaim the need for independent innovation, and when necessary, resort to outright copying—taking what is stolen and, if faced with resistance, resorting to theft. This reflects a bandit-like mentality, unrepentant despite the passage of time.

The distillation controversy within the CCP's AI sector represents a more advanced, digital, and covert iteration of these practices. Evidence from Anthropic and OpenAI points to the existence of fake account pools, credential theft, proxy networks, cross-session attacks, and even the forwarding of genuine customer requests to American models. This situation is not merely a case of one-way technological plagiarism; it involves the unauthorized reverse leakage of user data.

Users of the Dark Side of the Moon think they are using the domestic Kimi, but in reality, their requests may be routed to Claude. Meanwhile, DeepSeek users analyzed internal documents, revealing that specifications and strategic goals ended up in American logs. This is a significant irony, as it suggests a world without thieves, where even thieves can project a sense of 'confidence'. Ironically, the leaked information included surveillance related to PLA facilities and the logic of the public security system. The actual performance is commendable, capable of winning battles, and the effort to tell the Chinese story has unexpectedly turned into one of the most sensational incidents of the year.

In Chen Yixin's article, the 'risk of large-scale theft and leakage' has, under the management of its own enterprise, resulted in self-inflicted harm. It is not the Americans setting traps from across the sea, but rather Chinese companies hastily sending themselves abroad. Political security slogans are rampant, yet they are caught red-handed while trying to exploit resources without cost, which directly undermines the authorities' credibility.

When issues arise, shifting the blame becomes a polished exit strategy. The Ministry of Commerce has labeled the American accusations as 'politicized and instrumentalized' and as 'technological hegemony', threatening countermeasures. Chen Yixin, on the other hand, attributes the source of the risk to the technology controls and monopolies of 'relevant countries'. This narrative aligns perfectly with the discourse surrounding chip sanctions and the trade war, suggesting that the issue lies not in a lack of original capabilities or regulatory violations, but rather in external suppression.

While distillation itself is a neutral process, the existence of industrial-scale fake accounts, traffic hijacking, and violations of service terms raises the question: are these normal business practices? American institutions have explicitly stated that these activities 'constitute the core of China's AI development strategy, rather than being supplementary', and that 'the government is likely aware of this'.

The Chinese Communist Party (CCP) has largely refrained from providing specific rebuttals to the evidence presented, resorting instead to denial and threats, with no public accountability measures in place. Chen Yixin's proposals to 'plug loopholes and hidden dangers' and 'issue safety risk prevention guidelines' appear more as attempts to divert attention and a case of closing the barn door after the horse has bolted. It was only after being caught that they recognized the issues of data leakage and insufficient user safety awareness, yet they shifted the blame to 'some domestic users' rather than addressing the systemic failures within the enterprises.

This uncomfortable mindset is rooted in the logic of the CCP's system. The Party has heavily politicized the AI industry, aiming for it to serve as a new battleground for great power competition, the primary arena for technological rivalry, and a crucial player in military transformation, all while demanding absolute loyalty to the Party's leadership. History has shown that movements reminiscent of the Great Leap Forward often lead to false prosperity and resource wastage. While plagiarism may yield short-term gains and reduce disparities, it fails to generate genuine technological advancements or foster long-term growth; when issues arise, shifting blame only exacerbates the international cooperation environment and deepens self-isolation.

In his article, Chen Yixin discusses the military applications of AI in the US-Israel-Iran conflict, stressing that whoever controls the algorithms holds the initiative on the battlefield. This underscores the CCP's perception of AI as a tool for power rather than as an innovative technology aimed at improving people's lives and advancing social progress. When distillation is employed to enhance military capabilities, when user data is inadvertently leaked, and when political security is prioritized over technological innovation, the notion of 'healthy and orderly development' devolves into mere rhetoric and empty promises.

True competition in AI is grounded in an open ecosystem, independent research, ongoing investment, and adherence to rules, rather than in pirate skirmishes and blame-shifting afterwards. The strategies employed by the Chinese Communist Party (Zhonggong) have already led to setbacks in the high-speed rail and semiconductor sectors, and the consequences will be even more severe in AI. This is because the Chinese Communist Party focuses solely on distilling the capabilities of large models, while lacking the courage to embrace the free institutional environment of the West. Chen Yixin speaks grandly about party control but subtly acknowledges leaks, which vividly illustrates this prisoner mentality: projecting an image of total control while feeling deeply insecure beneath the surface.

(First published by the People News)△