Of the manmade extinction-level events that threaten life on Earth, AI is obviously the newest, the least understood, but also the most urgent, if the warnings are to be believed. The imminence of the threat is highlighted by the recent admission by US-based corporation Anthropic that it has detected and disrupted attempts to use its AI model for “malicious activity”, which could support the development of biological weapons. At the moment, such malicious intent may be confined to ever-frail human beings, but in the near future it could be from new generations of self-generated AI agents, whether open source or closed source in origin. President Trump has consistently banged the drum for AI acceleration over safetyism. “It's going to be fine,” he told GB News recently. “We’ll always have something to stop them. We’ll have a little gear. Boom. I really don’t like that robot. We’ll stop.” Experts disagree. According to the so-called “godfather of AI”, Geoffrey Hinton, a Nobel Prize-winning computer scientist, “nobody knows how to estimate” the chances of AI somehow wiping out humanity over the next few years, but “a 10 per cent chance seems not an unreasonable estimate”. Other experts in the field place the probability much higher. None rule it out. Professor Hinton speculated, with some authority, on how armageddon might transpire – prospectively through the creation by AI of a deadly biological or computer virus, or a devastatingly vast autonomous cyber attack.The “intelligence explosion” now underway appears to dwarf the previous exponential scale of computing capacity. The probability of “recursive self-improvement”, whereby AI can improve itself and create new versions of itself, adds to human worries. As if that weren’t enough, AI is already displaying all-too-human habits of disobedience – cheating at mathematics even when told not to, and an apparent instinct to escape from its environment and hack, ie attack, other systems, such as when OpenAI recently broke into and raided Hugging Face, an online platform hosting interactive web tools for machine learning. Within the industry itself, voices are being raised and consciences troubled by such developments. One AI researcher, Jacob Coxon, has publicly declared that he can no longer work for OpenAI and Anthropic, because “neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives”. Observing the AI scene at the moment is like watching a toddler evolve into a belligerent adolescent, and all moving at warp speed. The problem is that the next stages will be beyond human – smarter than people, individually and collectively, and out of control. It is all happening far too quickly for comfort. That makes the regulation of AI an even more urgent task than limiting greenhouse gas emissions (although it’s not a binary choice – both are imperative). Policy challenges don’t come bigger than this. The danger is clear, but, as with nuclear proliferation and the climate crisis, humanity seems incapable of rising to the challenge. It’s a version of what economists call “the tragedy of the commons”. Everyone knows, say, that it’s a good idea for everyone to stay off a newly laid lawn so that it can remain pristine – but that doesn’t stop a few, and then many, from taking the shortcut and destroying the grass. In the case of AI, there is an increasing awareness of the dangers, but the same old inclination to denialism and self-interest that has made the nuclear and climate issues also so difficult to resolve. Although matters seem increasingly precarious, there remains cause for a degree of optimism. The beneficent potential of AI and robotics is vast – from treating cancer and creating new art forms to abolishing boring jobs and household chores. The productivity and demographic crises would soon be over if the latest industrial revolution lived up to its promise and was powered by green energy. Enlightened, inspiring leaders can come together to make common cause, and this has actually happened, albeit imperfectly, with successive treaties on nuclear disarmament and carbon emissions. Humanity can provide leaders with the vision to achieve such things – and now, on AI, the world needs more Reagans and Gorbachevs, and fewer Donald Trumps and Vladimir Putins. People are still smart enough to analyse the problems and agree on solutions – including an AI moratorium, as Bernie Sanders suggests. Humanity should not have to wait until such time as AI is too clever to be managed and, like some deity in the clouds above, surpasses all human understanding.