Fear, status and algorithms: Thucydides in Artificial Intelligence era

Translated by Carlotta Faraci

Today, on the 24^th of September, a meeting between Donald Trump and Xi Jinping is scheduled to take place in Washington. Artificial intelligence is one of the issues on the summit agenda and has already been raised in the preparatory talks: on the 20^th of September, in New York, US Treasury Secretary Scott Bessent and Chinese Vice-Prime Minister He Lifeng discussed a US proposal to establish a bilateral mechanism for notifying risks associated with AI. Generally speaking, the technological rivalry between Washington and Beijing is often portrayed as a race to determine who is ahead; however, the issue is not merely about who is ‘winning’: progress by the other side can also translate into an economic, military or strategic advantage and, for this very reason, prompt the competitor to in turn accelerate investment, strengthen its capabilities or introduce new restrictions. This creates a dynamic in which the actions of one power fuel the reactions of the other. At the same time, the more this competition increases uncertainty and the risks of escalation, the greater the interest in keeping certain channels of cooperation open emerges.

The fact that this competition involves much more than mere technical advantage is also evident from the way in which Washington and Beijing define leadership in AI: on one hand, the US plan on artificial intelligence explicitly sets out a race for global leadership, linking American supremacy to economic competitiveness, national security, infrastructure and the international dissemination of hardware, models and software; on the other hand, on the Chinese side, the “Global AI Governance Action Plan” of July 2025 adopts language focused on openness, technological cooperation, shared infrastructure and greater access to artificial intelligence for countries in the Global South. Although the two documents are of a different institutional nature, both demonstrate that leadership also entails the ability to influence the development, distribution and governance of technology.

To describe this dynamic as merely a new arms race is an oversimplification, because artificial intelligence cuts across scientific research, industry, markets and defense simultaneously, involving Nations, businesses and universities. Hence the concept of a ‘geopolitical race for innovation’, in which security, economic interests and status coexist with forms of collaboration that a traditional arms race would struggle to account for. It is precisely this overlap that makes it difficult to interpret the other side’s moves, given that a technology developed for commercial purposes can also acquire strategic applications.

In international relations, this mechanism is known as the security dilemma and occurs when measures taken by a Nation to enhance its own security end up increasing the perceived insecurity of others, thereby triggering new countermeasures even in the absence of any initial aggressive intent. In the case of artificial intelligence, the dual-use nature of technology further complicates the issue, as capabilities developed for civilian or commercial purposes can also yield strategic advantages. It therefore becomes more difficult to distinguish between an advancement aimed at enhancing one’s own technological capabilities and one capable of altering the security balance.

One example comes from the competition in generative artificial intelligence. The success of ChatGPT has made the US technological advantage more apparent in the Chinese sources analyzed, whilst at the same time fueling concerns regarding dependence on an ecosystem developed in the United States, information flows and internal security. China has responded to these concerns by combining greater regulation of generative artificial intelligence with the strengthening of its own technological capabilities. However, the progress made by Chinese models has in turn altered the American perception of the competition: with the emergence of DeepSeek, fears have grown regarding a possible narrowing of the US technological lead and the strategic implications of Chinese advances. The point, therefore, is not that every innovation is automatically perceived as a threat, but that progress by one side can alter the way the other assesses its own position and, consequently, trigger new competitive responses.

A similar dynamic also emerges in Chinese thinking on the military use of AI, where an analysis of 140 publications appearing between 1988 and 2022 shows that US developments have served as a constant benchmark for assessing Chinese capabilities. In several of these contributions, indeed, the American lead fuels the perception of a gap to be bridged and reinforces the urgency of accelerating the development of China’s own capabilities, making the United States not only the main competitor but also the benchmark against which to measure its own position and calibrate the pace of competition.

It is precisely this dynamic – in which the progress of one power alters the way the other assesses its own position – that makes the reference to Thucydides so useful. In his account of the Peloponnesian War, the Athenian historian does not, in fact, attribute the behavior of the powers solely to the balance of their capabilities, but also cites fear, self-interest and honor as motivations for political action. Carefully transposed to the present day, these categories allow us to shift our focus from a purely quantitative question — who is ahead in the technological race — to a broader issue, linked to the significance that each power attaches to the other’s progress.

This is also why Xi Jinping’s reference to the ‘Thucydides’ trap’ during his meeting with Trump in Beijing last May is relevant but not sufficient to explain the relationship between the United States and China. The term is now used to describe the risk of conflict between an emerging power and a dominant one, but it tends to focus attention primarily on the shift in the balance of power. In the case of artificial intelligence, however, the way in which that shift is interpreted also matters. The fear stems from uncertainty about the other’s capabilities and the fear of losing ground; interest centers on tangible resources such as computing power, semiconductors, infrastructure, markets and standards, whilst ‘honor’ – reinterpreted in the present through the prism of status – also concerns the position that technological leadership enables a nation to claim on the international stage. The advantage in AI thus takes on material, strategic and symbolic value, helping to explain why any shift in relative standing can produce effects that extend beyond the technology itself.

Competition does not, however, preclude the possibility of cooperating on certain specific risks. Precisely because the other party’s capabilities are difficult to assess and some applications of artificial intelligence may also have military implications, reducing uncertainty becomes a shared interest. This convergence remains, however, limited to risk management: Washington and Beijing may have an interest in improving communication and preventing misunderstandings without, for that reason, giving up the advantages associated with technological leadership.

The agreement reached during the talks on the 20^th of September regarding the formalization of a bilateral dialogue on AI does not, therefore, mark a retreat in the rivalry between the United States and China; rather, it shows how rivalry and containment measures can develop in parallel. Both powers continue, in fact, to have strong incentives to pursue technological leadership, but it is precisely the possibility that uncertainty may produce undesirable consequences that also create an interest in developing mechanisms capable of mitigating at least some of these effects.

It is in the light of this tension that the reference to Thucydides remains relevant, because technological competition does not necessarily lead to conflict but can fuel new competitive responses whenever uncertainty about the other party’s capabilities grows, even when both sides recognize the existence of common risks. Today’s meeting should be viewed within the context of this ambivalence, and any progress on the notification mechanism would not signal an armistice in the competition over artificial intelligence, but rather the opportunity to establish basic risk-management tools whilst the rivalry continues. The point, therefore, is not whether the United States and China will continue to compete, but whether they will succeed in preventing the uncertainty generated by this competition from itself becoming a source of escalation.

Mondo Internazionale APS - Riproduzione Riservata ® 2026