In an episode of the program "Scenarios" on Al Jazeera from 17 September 2026, experts described the paradox of the artificial intelligence race: all sides acknowledge the risks, but none is prepared to pay the price of slowing down alone. Companies fear losing the market and investors, while states fear ceding strategic advantage to rivals. This dynamic determines not only the pace of development but also the shape of future regulation.
The structural forces here are obvious. Competition between the United States and China has turned AI into an element of national security and military superiority, not merely a commercial product. Pressure from investors and the logic of the market compel companies to accelerate development even when they speak publicly about safety. The historical precedent — the nuclear arms race — shows that voluntary restraints work only when there is mutual trust and verification mechanisms, which do not yet exist in the field of AI.
The current moment reinforces this trend. In 2026, no major power is prepared to open its advanced models to international audit, fearing technology leaks. Proposals to create an analogue of the IAEA for AI run up against the problem of trust: the parties would have to share data and secrets they consider the key to economic and military advantage. As a result, regulation remains national and fragmented.
Hidden interests coincide among apparent adversaries. American and Chinese companies, as well as their governments, benefit from preserving uncertainty: it allows them to continue investment and development without rigid global restrictions. Third parties — such as European regulators — are trying to introduce rules, but their influence is limited, since companies can move development to less stringent jurisdictions.
The most likely outcome in the next two to three years is a continuation of the race with partial and scattered measures: national laws, voluntary industry standards, and technical "safeguard" solutions such as the isolation of critical systems. A full-fledged international treaty is unlikely because of the asymmetry of information and the fear of losing an advantage. This scenario is confirmed by the logic of incentives: no one wants to be the first to slow down until an obvious and costly threat emerges.
The two strongest counterarguments are the possibility of a sudden breakthrough in safety or a major accident that would force immediate action. However, the first requires unprecedented cooperation, which the history of such races does not demonstrate, while the second, according to experts' assessments, does not yet appear inevitable in the coming months. The forecast will break down if the United States and China suddenly agree on minimum verification standards, or if a series of AI incidents in critical infrastructure occurs as early as 2027.
The key indicator to watch in the coming 4–8 weeks is the decisions of national regulators in the United States and the European Union on mandatory testing and reporting for advanced models, as well as any joint statements by Washington and Beijing on AI. If these steps remain declaratory or national, the forecast of fragmented regulation will be confirmed.
The takeaway for the reader: track not the loud statements about risks, but the concrete regulatory acts and investment decisions of companies — they are what will show the real trajectory of the race.
