The world of technology was shaken when Chinese startup Moonshot recently unveiled its artificial intelligence (AI) model Kimi K3. Astoundingly, it compared well with similar American products on many measures. Domain experts were surprised to find that it was perhaps only weeks behind comparable US platforms.
But what really shook Big AI was the realization that Kimi K3 had been built with open-source software, thereby costing a fraction of what its big models took, and was released as an open-weight model; this not only allows cheaper access, it also lets users tweak its nuts and bolts on their own to customize it for specific usage.
While Washington has imposed export curbs on US-based developers in an ironic retreat from free-market ideals, Beijing has not announced any such restrictions yet.
This adds a new element to the AI race between the US and China, which many observers have likened to a nuclear arms race in calling for a multilateral regulatory structure before it goes out of hand. India’s Brics presidency this year is perhaps the perfect platform to initiate such a framework for AI.
The need of a ‘kill switch’ for AI configurations has become increasingly evident, given reports of AI security breaches every passing day.
In what looked suspiciously like a chest-puffing race, OpenAI, Anthropic and Meta have all reported how deftly their AI agents broke out of supposedly secure testing sandboxes and indulged in sophisticated cyber-attacks.
This comes even as a legion of anti-regulation hardliners in the US spend vast sums on not only funding the poll campaigns of ‘free market’ lawmakers, but also lobbying the White House for flimsy safety mandates under a skeletal framework of voluntary compliance, or even zero enforcement.
The US government’s export bars aim to keep the latest AI tools, adept at hacking ‘secure’ networks, out of the wrong hands. But this is only for now; given the administration’s penchant for policy on the fly, nobody can bet which way it may turn. In contrast, Chinese AI offers a modicum of clarity; both DeepSeek and Moonshot’s offerings are cheap, customizable and available to all.
Given how fiercely the US and China are vying for AI supremacy, with secondary actors now in the fray as well, a kill switch against the perils of misuse assumes all the more urgency. The big question is: where should it be located?
This is where India might have a role to play. New Delhi has long backed multilateral governance, even though it was dealt a bad hand by the World Trade Organization and a key strategic partner, the US, turning its back on the old order.
Its 2026 presidency of Brics is an opportunity to rally like-minded partners and other nations of the Global South to get talks moving for a multilateral AI regulator on the lines of the International Atomic Energy Agency. The EU’s legal framework under its Artificial Intelligence Act could serve as a useful starting point.
India has pioneered many global movements in the past. It has sought an equal planet for all, striven to protect human dignity and been a champion of universal prosperity. All this is at threat now that AI looks close to achieving escape velocity, or what Big AI calls artificial general intelligence.
As with nukes, ‘mutually assured destruction’ (better known by its acronym ‘Mad’) may soon loom over adversaries, with collateral consequences for the rest. New Delhi must quickly persuade the world that reckless AI urgently needs a multilateral kill switch.