A recent warning by the leading intelligence chiefs pertaining to the growing threat of Artificial Intelligence-led cyber wars is a reminder that a world grappling with multiple, ongoing conflicts must be ready for an even more unpredictable future. The Five Eyes alliance, consisting of the United States of America, the United Kingdom, Canada, Australia and New Zealand, has cautioned that Western nations are in a race against time to fortify themselves against AI-engineered cyber attacks, and that their current defensive capabilities might become inadequate within months. The warning comes close on the heels of the US government banning the leading AI firm, Anthropic, from allowing non-Americans access to its most advanced models. It also comes amid sharpening geopolitical tensions over the AI race, involving not just the US and traditional adversaries like China and Russia but also Washington's fissures with historical allies like Europe. All of these rapid shifts are undergirded by an even more alarming set of fears surrounding a central question: given the speed at which AI is advancing, can it be controlled even by those who might unleash its power for cyber operations against other countries, companies or organisations?

That is not just a theoretical worry anymore. Experts warn that a situation where two AI-powered cyber operations come up against each other is no longer far-fetched. They could each learn, adapt and try to outmanoeuvre the other in real time, changing strategies and targets, and escalating their cyber war at a pace that might become hard to stop through manual intervention. This is especially a concern given the rupture, more broadly, in global rules and trust. Europe knows that the US spies on it. The US knows that Israel spies on it. As countries like the US try to limit access to their more advanced AI products, other nations will develop their indigenous versions — China is already a peer competitor, with its own sophisticated AI firms and models. More and more nations will opt for opacity around their AI programmes. India, too, must prioritise building its homegrown capacity to develop cutting-edge AI models. Some efforts have been made, but need to be speeded and scaled up. The world has seen such a crisis before, when nations clandestinely raced to develop atomic weapons in a 20th-century climate of mistrust. Lessons must be learnt from that episode: a global compact on AI regulations is critical, and unlike with atomic weapons, it must not foster a system of haves and have-nots.