When Anthropic announced its Claude Mythos Preview on April 7, 2026, the San Francisco-based AI lab made an unprecedented decision: the model was deemed too powerful to release publicly due to severe cybersecurity risks. Access was restricted to a small, vetted group of US technology companies and government agencies. In China, the initial reaction was largely dismissive. Discussions on the popular Q&A platform Zhihu, which garnered nearly 2 million views, were dominated by accusations that Anthropic was merely generating hype to attract investment. However, as the implications of the model’s capabilities have become clearer, that initial scepticism has rapidly evolved into genuine concern among Chinese cybersecurity experts and policymakers.
The turning point came following reports of an emergency meeting between the US Treasury, the Federal Reserve, and major US banks to discuss the potential threats posed by Mythos. This high-level response signaled that the model’s capabilities were not merely marketing spin. Shortly after, Beijing commented directly on Mythos for the first time via Yuyuan Tantian, a social media account affiliated with state broadcaster CCTV. The commentary acknowledged that Mythos demonstrated “unprecedented cyberattack capabilities,” marking a significant shift in the official narrative.
The Cybersecurity Vulnerability Gap
The core of the concern in China centers on the country’s vast and complex telecoms and internet infrastructure. Because Chinese companies have been explicitly excluded from accessing Mythos, domestic cybersecurity teams are unable to study the model’s attack vectors or develop countermeasures. A Beijing-based independent researcher, speaking anonymously to the South China Morning Post, articulated the growing anxiety: “We are entering an era of AI-driven cyberattacks, but we are in a disadvantaged position as our companies have been excluded from accessing the model.”
This exclusion is not accidental. Anthropic CEO Dario Amodei has been a vocal proponent of stringent US export controls on AI technology to China. Among major US AI companies, Anthropic maintains some of the strictest access restrictions, including real-name verification processes that require government ID checks to prevent unauthorized use by Chinese entities. This hardline stance has effectively created an asymmetric security environment where US defenders can study the latest AI-driven threats, while Chinese defenders are left in the dark.
The potential consequences of this asymmetry are severe. China plays an outsized role in global supply chains, meaning a successful cyberattack on its IT or logistics platforms could trigger widespread disruptions. S. Alex Yang, a professor of supply-chain finance at London Business School, highlighted the specific vulnerability of older operational technology (OT) systems. “Older OT systems are a particular concern, as many were designed decades ago for reliability and continuity, not for today’s cybersecurity environment,” Yang noted. The fear is that an advanced AI like Mythos could identify and exploit vulnerabilities in these legacy systems far faster than human hackers.
The Push for Domestic Alternatives
While some legal and cybersecurity professionals in China report not yet receiving panicked queries from clients, the broader industry is already feeling the impact. Tony O’Sullivan, CEO of UK internet services provider RETN, which operates terrestrial cable infrastructure in China, stated that his company is already seeing AI systems being used maliciously to automate high-speed DDoS attacks. This real-world evidence of AI-weaponization underscores the urgency of the threat.
In response to the Mythos challenge, Chinese state media has begun emphasizing the need for domestic AI breakthroughs. The CCTV commentary that acknowledged Mythos’s capabilities also pointed to DeepSeek’s latest model and its collaboration with Huawei as “a notable breakthrough.”
(Related: DeepSeek Releases V4 with 1.6 Trillion Parameters and Huawei’s Full Support)
This framing suggests that Beijing views the development of indigenous, frontier-level AI models not just as an economic imperative, but as a critical national security requirement. “When we discuss Chinese AI companies, we are not just talking about the companies themselves, but also the industry and the development model they represent,” the CCTV commentary stated.
The Mythos episode highlights a growing divergence in the global AI landscape. As US companies develop increasingly powerful models and restrict access on security grounds, China is forced to accelerate its own AI development to avoid falling behind in both capabilities and defensive knowledge. The era of open, global AI research is rapidly closing, replaced by a new paradigm of guarded, national-security-driven AI development. This shift is particularly acute for Chinese enterprises that rely on international technology partnerships. As the US government and leading AI labs increasingly view advanced models as dual-use technologies with significant military and cybersecurity implications, the barriers to entry for Chinese firms will only rise. The immediate challenge for Beijing is not just developing a model capable of matching Mythos, but doing so before the existing cybersecurity vulnerability gap can be exploited.