How the AI arms race traps us all on an upgrade treadmill

The Relentless Machine: How AI's Upgrade Treadmill Traps Millions in Perpetual Adaptation

From scam alerts in Beijing community centres to emergency retraining programmes in London offices, the accelerating pace of AI disruption has created something unprecedented: a global treadmill where standing still means falling behind. This isn't an accidental byproduct of technological progress. It's the engineered outcome of a new commercial model that profits from perpetual obsolescence.

The numbers paint a stark picture. AI-enabled adversaries increased operations by 89% year-over-year in 2025, whilst average cybercrime breakout time plummeted to just 29 minutes. For millions caught in this acceleration, the choice is binary: adapt constantly or risk irrelevance.

China's High-Speed Preview of Tomorrow

Tencent's recent decision to open-source two world-class^✦ AI platforms offers a glimpse into this relentless future. Hunyuan-MT handles internet slang with uncanny precision, whilst Hunyuan-Voyager generates 3D worlds from single images. These aren't just technical achievements; they're existential challenges wrapped in accessibility.

In February, Hong Kong's Arup branch lost HK$200 million when deepfake fraudsters impersonated senior executives during a video call. The incident catalysed widespread anxiety, but it also revealed something more troubling: the burden of defence increasingly falls on individuals.

Beijing's pensioners now attend weekly workshops to identify deepfake calls. Professional translators in Shanghai race to master neural models that threaten their livelihoods. The pattern is consistent across sectors and age groups. China's AI race has become a societal-wide sprint where pause equals peril.

"Countries think they must control AI before it controls them," observes the Atlantic Council in its 2026 geopolitics analysis. This driving force behind sovereign AI^✦ announcements has sparked massive infrastructure investments, including the US's $500 billion Stargate project.

The Western Tremor: Slower but Inevitable

Western societies experience similar disruption at a more manageable pace, yet the fundamental dynamics remain unchanged. Generative AI adoption across Asia reveals how quickly the treadmill accelerates once it gains momentum.

Anthropic's recent decision to cut access for Chinese firms illustrates how geopolitics compounds individual burden. Chinese companies that had integrated Anthropic's models must now pivot^✦ to domestic alternatives. Intended as strategic containment, the move has inadvertently fuelled the arms race it sought to prevent.

A Pew Research Centre survey found that more than half of US workers worry about AI's impact on their careers. Yet this anxiety has spawned an entire industry. Coursera reports surging enrolments in AI-related courses, whilst organisations like Age UK now offer scam-prevention programmes as public necessities.

The paradox is striking. Meta faces criticism for enabling misinformation on its platforms, yet simultaneously offers AI literacy tools. The harm is systemically created whilst mitigation costs are individually borne.

The Individualisation of Technological Risk

Perhaps the most insidious aspect of this treadmill is how it reframes systemic problems as personal failings. Workers who struggle to keep pace aren't victims of an accelerating system; they're insufficiently adaptable.

"Breakout time reflects how quickly intrusion methods are evolving, and security teams must operate at greater speed to respond effectively to modern threats," explains Adam Meyers, head of counter adversary operations at CrowdStrike.

This statement, whilst technically accurate, reveals the broader assumption underlying our current approach. The solution to technological acceleration is always more acceleration, never deceleration or stability.

For professionals navigating Asia's AI literacy race, the new reality demands constant vigilance: scan repositories, test models, enrol in courses, or risk obsolescence. What masquerades as professional development is often survival training against systems optimised to outpace human adaptation.

The evidence suggests we're trapped in what economists call a Red Queen scenario: constant running just to maintain position. Yet unlike biological evolution, this arms race is artificially maintained by business models that profit from disruption.

Beyond Individual Solutions

The upgrade treadmill isn't inevitable; it's engineered. AI's impact on Asia's farming communities demonstrates how technological solutions often create more problems than they solve when deployed without considering human adaptation limits.

Several alternative approaches deserve consideration:

• Stability-first development that prioritises human adaptation timelines over market speed
• Collective responsibility models that distribute adaptation costs across organisations rather than individuals
• Regulatory frameworks that mandate adaptation support alongside technological deployment
• Open-source initiatives that reduce learning curves rather than steepening them
• Industry standards that slow deployment rates to sustainable levels

These aren't utopian proposals but practical alternatives to our current trajectory. The question isn't whether we can step off the treadmill, but whether we choose to.

Why does the AI upgrade treadmill seem unavoidable?

Business models profit from constant disruption and obsolescence. Companies create anxiety about falling behind, then sell solutions to that anxiety. This cycle becomes self-reinforcing as competitive pressure forces participation.

How does China's approach differ from Western strategies?

China prioritises speed and open-source adoption, accepting instability for competitive advantage. Western approaches emphasise individual training and corporate responsibility, whilst Europe focuses on regulatory protection. Neither fully addresses systemic issues.

What are the hidden costs of constant adaptation?

Beyond financial expenses, constant upskilling creates chronic stress, reduces deep expertise development, and shifts resources from productive work to defensive learning. Mental health impacts are significant but rarely quantified.

Can regulation slow the treadmill effectively?

Regulation can establish minimum adaptation timeframes and mandate corporate responsibility for worker retraining. However, global competition means unilateral action risks competitive disadvantage unless coordinated internationally.

What would stability-first AI development look like?

It would prioritise backward compatibility, longer deployment cycles, comprehensive training programmes, and human adaptation timelines over market speed. Companies would measure success by sustainable adoption rates rather than disruption metrics.