With China and America in an AI-catapulted space race, the world needs a global treaty on AI warfare

In the last week of December, Beijing notified the UN’s International Telecom Union of its plan to put over 200,000 satellites in orbit within a decade. Its game, it seems, is to saturate the whole planet with satcom coverage and make US-based SpaceX’s Starlink constellation of almost 10,000 low-earth orbiters look skeletal. Clearly, artificial intelligence (AI) is not the only tech race China is in a hurry to win.

From the viewpoint of global geo-rivalry, satcom and AI are twin frontiers. Over in the US, AI startup Anthropic is reported to have had a run-in with the Pentagon over curbs on the use of its Claude AI tools, which other reports suggest were used by US forces for a regime-switch in Venezuela.

AI paired with satcom can enable armed action—think of AI-run drone swarms armed with live data feeds across the globe—that might tip the balance of power one way or another. This may also explain why Elon Musk’s SpaceX snapped up xAI, his AI venture.

We can expect a blitz of space launches by archrivals America and China, one that grabs almost all low-earth orbits before others get a chance. How this shapes up matters to the world, of course, but especially to a space power like India. As defence minister Rajnath Singh said in the context of advanced technologies, “We must stay ahead of the curve.”

Is China bluffing? Without a reusable rocket like SpaceX’s fabled Falcon 9, which places satellites in orbit and returns home, Beijing’s plan seems to stretch credibility.

But then, that’s where its Chang Zheng 9 plan may prove handy. If it works out, not only will this space vehicle be ready for round trips by the early 2030s, it would haul 150 tonnes of stuff into low-earth orbit with each go, more than six times what Falcon 9 does. That could make China’s eye-popper of a plan plausible in terms of both bulk and cost.

As Musk has shown, even space launches are now about economies of scale—which is why SpaceX’s cost-per-tonne has reputedly dropped below Isro’s. While Isro specializes in low-cost single shots, it is working on vehicle reusability too.

For the sake of our satcom sovereignty—or self-sufficiency—this programme may need to be sped up. After all, relatively close orbital slots could turn scarce as satellite constellations begin to act as ‘the nervous system’ of AI-enabled warfare. Closeness offers an edge on signal quality for data-rich beams that need to dodge jammers.

If the role played in modern combat by such networks is hard to overstate, so is that of AI itself. Speedy responses are at a premium in battle. For most of history, combatants have had to identify a threat, decide on a course of action and then respond. An AI system that’s given autonomy could snip this loop into a split-second. Given the fallibility of AI, nuclear triggers need not be in play for this to prove perilous. But then, that’s the trouble with any arms race: restraint could lose traction as the stakes rise and turn the contest into a spiral of speed.

As an imperative, thus, we should insist on human oversight of all armed action, with decisions kept under the control of actors with proven judgement. What’s AI-aided must not become AI-led. No doubt, we must rush to keep up with tech advances. But let’s also advocate a global treaty on AI warfare that could make the world safer for everyone.