Milano Cortina closes on a high-tech note

The Milano Cortina Winter Olympics closed on 22 February, and the afterglow of the fortnight is being measured in more than medals. The Games will be remembered, organisers and observers agree, as the most technologically integrated edition yet, the first in which a foundation model was embedded in the core running of the event. The official Olympic model, built on the Qwen large language model and announced days before the opening ceremony, handled scheduling and resource coordination, translated between dozens of languages, and sat behind the assistance offered to athletes, delegations and officials. It did not merely support the Games; it was, for seventeen days, part of their nervous system.

The technology was visible to spectators in forms that became quietly routine. Broadcasts relied on drone cameras and AI-assisted systems that could break a downhill run into a clear sequence of positions, or reconstruct a figure skater’s trajectory in three dimensions within seconds of the landing. Real-time replays stopped treating complex sports as mysteries; a biathlon race, with its alternation of skiing and shooting and penalties, could be explained to a first-time viewer on screen as it happened. The tools did not replace the human broadcast, and the coverage was still anchored by commentators, but the machines added a layer of legibility that previous Games lacked.

For the athletes and the delegations, the model worked in the background: a schedule changed, a venue was distant, a document existed only in another language, and the system absorbed the friction that used to consume hours. The International Olympic Committee had framed the deployment as a test at maximum scale, and by the closing ceremony the test had, by the available evidence, been passed: the Games ran, the operations held, and the technology earned no headlines of its own, which is the highest compliment an infrastructure can receive.

The retrospective is not entirely unqualified. The fortnight also demonstrated how much of the Olympic experience now depends on systems whose inner workings few people understand, and the organisers’ claims of efficiency will deserve independent auditing rather than applause. Data use at scale raises questions about privacy and consent that a few weeks of smooth operation do not answer. And the technological framing risks obscuring the sporting story, the athletes whose years of training produced performances that owed nothing to any model. The right lesson from Milano Cortina is that AI can carry infrastructure well, and that infrastructure is not the event.

Still, the direction is settled. The IOC has now run a Games with a foundation model at its core, and no future edition will be organised as though that had never happened. The model was built in China, deployed under international governance, and used by participants from more than ninety delegations, a small, concrete example of a globalised technology supply chain working under public scrutiny. The flame has been passed on, and with it the knowledge that the next Games will be expected to run at least this well, with the same invisible machinery, at even larger scale.