Why we must take seriously China's mastery and misuse of AI espionage

In William Gibson’s science fiction novel Neuromancer, artificial intelligence is depicted as being used for espionage and to manipulate international relations. The novel revolves around a washed-up computer hacker hired by a mysterious employer to pull off the ultimate hack. In the process, he encounters AIs that manipulate individuals and events to serve their ends, subtly influencing global power structures.

While we haven’t reached the dystopian future of AI depicted in ‘Neuromancer’, where artificial intelligence becomes a direct threat to human existence, the world is witnessing the early stages of AI’s potential for harm. Countries have begun to harness AI for espionage, sowing discord in foreign nations and inciting political unrest. These actions mark the subtle beginnings of AI’s potential to manipulate and destabilise international relations, and China is at the forefront of this.

This week’s Microsoft Threat Intelligence report starkly illustrates this trend. It details how China has been actively using AI-generated content to exacerbate divisions within the US, Asia Pacific region. (including Japan, Taiwan and South Korea), and even in India. With AI, disinformation can be more effectively tailored and disseminated to stir conflicts, influence public opinion, and even tamper with the democratic process.

As noted in the report, with upcoming elections in India, South Korea, and the United States, there is a significant concern that Chinese, and to some extent North Korean, cyber and influence operations will intensify. These actors will likely leverage AI capabilities to target the electoral processes, aiming to weaken trust in democratic institutions and influence election outcomes.

Chinese state-sponsored hackers increasingly employ sophisticated AI-powered tools to enhance their cyber espionage capabilities, targeting critical infrastructure, government agencies, and private companies. These advanced persistent threats (APTs) exploit vulnerabilities in networks and systems to gain unauthorised access, exfiltrate sensitive data, and potentially disrupt essential services. Moreover, China’s cyber actors are adept at using AI algorithms to analyse vast amounts of stolen data, identify high-value targets, and craft highly personalised and convincing phishing attacks.

China-based espionage groups are intensifying geopolitical tensions in the South China Sea through sophisticated cyber espionage activities targeting strategic partners and rivals alike. These groups, identified by Microsoft Threat Intelligence as Gingham Typhoon, Flax Typhoon, Granite Typhoon, and Raspberry Typhoon, have been actively engaging in cyber operations that reflect China’s broad strategic objectives in the region. Gingham Typhoon, in particular, has been the most active actor, targeting international organisations, government entities, and the IT sector across nearly every South Pacific Island country.

This includes complex phishing campaigns aimed at vocal critics of the Chinese government as well as diplomatic allies of China, highlighting the dual goals of extending global influence and gathering intelligence. The espionage efforts are not limited to political and military targets but also encompass economic partners, as seen in the large-scale targeting of multinational organisations in Papua New Guinea, a nation benefiting from China’s Belt and Road Initiative (BRI) projects.

Moreover, the focus of these espionage activities extends to entities related to the South China Sea, where China-based threat actors opportunistically compromised government and telecommunications victims within the Association of Southeast Asian Nations (ASEAN). This targeting behaviour is particularly pronounced in the context of US military drills in the region, with Raspberry Typhoon successfully targeting military and executive entities in Indonesia and a Malaysian maritime system ahead of a multilateral naval exercise involving Indonesia, China, and the United States.

Similarly, Flax Typhoon targeted entities related to US-Philippines military exercises, while Granite Typhoon compromised telecommunication entities across Indonesia, Malaysia, the Philippines, Cambodia, and Taiwan.

Chinese espionage groups, such as Storm-0062 and Volt Typhoon, have notably escalated tensions in the United States through targeted cyber activities against military and critical infrastructure sectors. Storm-0062’s focus on aerospace, defence, and natural resources, coupled with Volt Typhoon’s infiltration of critical infrastructure networks, reflects a strategic effort to undermine US national security. These actions not only compromise the integrity of vital sectors but also raise alarm about the potential access these groups have to sensitive information, thereby stoking fears and mistrust within the US defence community and beyond.

In Taiwan, the influence operations led by Storm-1376, a Chinese Communist Party-linked actor, have injected additional strain into already tense cross-strait relations. By utilizing advanced AI to fabricate endorsements and spread disinformation during critical election periods, these actors have sought to manipulate public perception and political dynamics in Taiwan. The deployment of AI-generated content, including fake endorsements and misleading narratives, represents a sophisticated escalation in the tactics used to influence Taiwan’s political landscape, exacerbating tensions between Taiwan and China.

Japan and South Korea have also been targets of Chinese influence operations, with Storm-1376 amplifying controversies and stoking discord within and between these nations. In Japan, the group spread fear and misinformation regarding the disposal of Fukushima’s treated radioactive wastewater, challenging scientific assessments and sowing doubt about the safety and intentions behind the disposal. Similarly, in South Korea, the group capitalised on environmental and diplomatic concerns, using localised content to amplify protests and criticisms against the Japanese government. These actions not only exacerbate regional tensions but also aim to undermine trust in governmental and international regulatory bodies.

Moreover, the spread of conspiratorial narratives, such as the claim that the U.S. government used a “weather weapon” in Hawaii, alongside aggressive messaging campaigns in South Korea and misinformation surrounding the Kentucky train derailment, illustrates a broader strategy by Chinese espionage groups. By exploiting and amplifying regional and domestic issues, these groups aim to foster distrust, deepen societal divisions, and weaken the coherence and international standing of the United States, Taiwan, Japan, and South Korea. This multifaceted approach to stoking tensions reveals a complex and persistent threat that these nations must address collectively to safeguard their security and democratic processes.

To enhance the technical sophistication of India’s response to AI-driven cyber espionage, NTRO could leverage state-of-the-art machine learning models and anomaly detection algorithms to improve threat detection capabilities. By adopting deep learning techniques, such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs), for analysing network traffic and user behaviour, NTRO can identify subtle patterns indicative of malicious activities that traditional systems might overlook. These advanced models can be trained on vast datasets of cyber incidents to accurately predict and detect espionage activities. Furthermore, the integration of adversarial machine learning can help NTRO’s systems to anticipate and counteract evasion tactics employed by AI-powered cyber threats.

DARPA’s initiatives, like the CHASE program, utilise big data analytics and machine learning to automate the detection of cyber threats on a large scale. NTRO can adopt similar methodologies, employing scalable data processing platforms and real-time analytics to continuously monitor and analyse cyber threats. This approach would enable the rapid identification of anomalous behaviours and potential cyber espionage activities, facilitating preemptive actions against such threats.

To tackle AI-generated disinformation, techniques like digital watermarking and blockchain can be employed to authenticate content and trace its origin. Advanced AI detection tools, which analyse inconsistencies in image or audio files, can be used to spot and flag deepfakes and synthetic media. These tools often utilise feature extraction methods and classification algorithms to differentiate between genuine and manipulated content. Developing and implementing these AI-driven detection tools requires a collaborative effort between government agencies, academia, and the tech industry to continuously refine algorithms and adapt to evolving disinformation tactics.

To address the challenge of AI-generated content, platforms where such content is shared must take on greater responsibility and adopt more robust measures. This necessitates a multi-layered approach involving technical, regulatory, and collaborative strategies to ensure the integrity and trustworthiness of the information disseminated. Technically, platforms need to implement advanced detection systems that can identify AI-generated content with high accuracy.

These systems should leverage the latest advancements in machine learning, such as natural language processing (NLP) and image recognition algorithms, to analyse and detect patterns or anomalies characteristic of synthetic media. For instance, employing deep learning techniques like Generative Adversarial Networks (GANs) can help discern between real and AI-generated images or videos by detecting subtle discrepancies that are typically invisible to the human eye. Regulations should make platforms more responsible for the content being disseminated. Further, there is a need for more collaboration between academia, platforms and government technical intelligence agencies to deal with AI-generated content.

Ironically, one should invoke Sun Tzu’s The Art of War here. “Know the enemy and know yourself, and you can fight a hundred battles without disaster,” Sun Tzu wrote, signifying the importance of understanding both the capabilities of AI and the tactics of adversaries like China. To combat AI-induced espionage and division, nations must embody Sun Tzu’s principles by developing a deep knowledge of AI’s potential for both creation and deception, ensuring preparedness for any cyber threats. Similar to Sun Tzu’s emphasis on planning and adaptation, strategic foresight must guide the defence against these cyber incursions.

The author (X: @adityasinha004) is Officer on Special Duty, Research, Economic Advisory Council to the Prime Minister of India. Views expressed in the above piece are personal and solely that of the author. They do not necessarily reflect Firstpost’s views.