Cina and USA in the AI race: vulnerabilities in water systems and use of American AI models

This week, two significant developments in the landscape of cybersecurity and artificial intelligence have attracted global attention: the use of U.S. AI models by Chinese military research units and the expansion of cyberattacks on U.S. water systems.

Chinese military units use U.S. AI models for defense

Researchers have published a report accusing Chinese military units of using U.S. AI models to train smaller, more portable defense systems. The report, based on an analysis of over eighty Chinese academic articles and patents, reveals that Chinese military units are employing a technique known as "model distillation." This technique allows the use of more powerful AI systems to train smaller, specialized models that can be rapidly deployed without requiring high computational power.

The researchers noted that Chinese defense institutions see these advanced models as a source of technical insight and a way to bridge the AI gap with the United States. Sunny Cheung, a Jamestown fellow who analyzed over sixty articles published by the Chinese army on AI, observed: "These articles show that military researchers linked to China are trying to transfer that expensive, proprietary reasoning from Western models into smaller systems that they can control and deploy locally."

For example, in an article published by Unit 96941 of the People's Liberation Army (PLA), the group highlighted the use of GPT-3.5 to process sensitive military source code. Chinese researchers emphasized that third-party models are not suitable for handling classified information and that GPT-3.5 can summarize software code and train national models.

Implications for U.S.-China AI Talks

This research emerges as the United States and China plan to meet in September to discuss the risks posed by frontier AI models. Treasury Secretary Scott Bessent commented: "We are finding watermarks of our American language models on many of the Chinese models, and this is unacceptable."

Furthermore, the incident could push the United States to strengthen controls on the export of AI models. The Biden administration had implemented an AI dissemination framework in January 2025, but the Trump administration revoked it in May 2025, stating that it would "stifle American innovation and burden companies with onerous new regulatory requirements."

Challenges in the Global AI Race

This research represents one of the biggest challenges in the global AI race. While the United States has limited Chinese access to advanced semiconductors, preventing access to frontier AI models is much more difficult. Techniques like model distillation allow researchers to extract capabilities from larger proprietary models and transfer them into smaller systems that can be operated nationally with much less computational power.

The findings will likely increase demands on the federal government to increase controls on frontier AI models. More broadly, the report highlights a growing reality for AI competition. As frontier AI models become increasingly tied to military assets, policymakers will need to assess who develops these models, who has access to them, and how they are used both nationally and internationally.

Cyberattacks on U.S. water systems continue to grow

In recent days, reports have emerged indicating that water providers in several states have been hit by a coordinated cyberattack. More specifically, water providers in at least seven states have been affected. The attacks have caused operational disruptions, forcing some utilities to switch to manual operations to ensure the safety of the water supply.

The full extent of the damage and the identities of the perpetrators are still under investigation. However, cybersecurity experts warn that such attacks could become more frequent as hackers target critical infrastructure.

The economic and social impact of attacks on water infrastructure

Cyberattacks on water infrastructure have a significant impact not only on national security but also on the economy and society. Operational disruptions can cause substantial financial losses for water utilities, which must deal with repair costs and potential fines for regulatory violations. Additionally, the need to switch to manual operations can slow down the production and distribution of drinking water, affecting the quality of life for citizens.

From a social perspective, cyberattacks can generate panic and insecurity among the population. The awareness that critical infrastructures can be vulnerable to external threats can erode trust in institutions and public services. This can have long-term effects on social cohesion and political stability.

The challenges of international cooperation

The transnational nature of cyberattacks requires international cooperation to effectively address the threats. However, cooperation between the United States and China, particularly regarding cybersecurity, is complex due to political tensions and strategic divergences. The United States has recently imposed restrictions on the export of advanced technologies to China, which could limit the capacity for collaboration in cybersecurity.

Additionally, the lack of a clear international legal framework for managing cyberattacks makes it difficult to adopt coordinated measures. This creates a void that malicious actors can exploit, making it even more urgent to engage in constructive dialogue among nations.

Technological innovations for the defense of critical infrastructures

To counter cyber threats, critical infrastructures must adopt advanced cybersecurity technologies. Among these, artificial intelligence can play a crucial role in detecting and responding quickly to attacks. AI systems can analyze large amounts of data in real time, identifying anomalies and potential vulnerabilities before they can be exploited.

Furthermore, the implementation of secure communication networks and the use of advanced authentication protocols can improve the resilience of infrastructures against cyberattacks. Water utilities should invest in technologies that enable a rapid and coordinated response to security incidents, minimizing the impact on operations.

The role of the private sector in cybersecurity

Technology companies and IT service providers have a fundamental role in protecting critical infrastructures. Collaborations between the public and private sectors can accelerate the development and implementation of advanced security solutions. Companies can contribute with their expertise in managing large networks and developing secure software.

Additionally, partnerships between governments and technology companies can facilitate the exchange of information and best practices to improve cybersecurity. This integrated approach is essential to address the complex and evolving threats that characterize the cybersecurity landscape.

The need for a long-term strategy

To ensure the security of critical infrastructures, it is essential to adopt a long-term strategy that includes continuous investments in research and development, staff training, and updating regulations. Institutions must be ready to quickly adapt to new threats and implement preventive measures.

An effective strategy also requires public awareness and the promotion of safe online behaviors. Educating citizens about the importance of cybersecurity can help create a culture of prevention and resilience against cyberattacks.

Cyberattacks on water infrastructure represent a serious and growing threat to national security and social well-being. Addressing this challenge requires a multifaceted approach involving governments, companies, and citizens. Investing in advanced technologies, promoting international cooperation, and adopting long-term strategies are essential steps to protect critical infrastructures and ensure the safety of drinking water for all.