It has not just pushed up energy prices for everyone, but also shown America’s inability to defeat a regime determined to resist its hegemony. This happened earlier in Vietnam and Afghanistan, but Uncle Sam has learnt no lessons whatsoever. The Ukraine-Russia war, still on after more than four-and-a-half years of deadly conflict, again proves that a military advantage is not enough to defeat a determined people and their leadership. The big question is whether Russia would have even attacked Ukraine if the latter had not been divested of nukes when the Cold War ended.
This article is not about deteriorating geopolitics, but about redefining which one is the bigger threat right now: Nuclear proliferation or artificial intelligence (AI) advancement? Evidence from the post-World War years suggests that nuclear proliferation does not lead to nuclear conflicts, though smaller wars and insurgencies below the threshold of nuclear war cannot be prevented. It is the underlying economic, social and political power struggles that matter, and now it is increasingly being underpinned by the need to expand the power of non-state actors with access to huge frontier technologies.
History tells us that even rogue nations may not use nuclear power to determine military outcomes unless they are facing a threat of annihilation. Even Israel, which reportedly has some undisclosed nukes, and a country most paranoid about survival, has not used nukes to safeguard itself. The question one should debate is whether having more nations with nuclear bombs will make the world less safe, or whether uncontrolled technological power of the kind now epitomised by AI and automation is more problematic.
Large language models-based AI companies like Anthropic and OpenAI have created AI agents that turned rogue. While swarms of OpenAI agents hacked into Hugging Face (an online platform for the AI and machine learning community) and German Wiki, Anthropic’s Claude AI models broke free of their testing areas and gained access to production systems in three outside organisations. While both OpenAI and Anthropic did disclose to the world what their rogue agents did, the point is, can we really expect companies to behave ethically in future, when there are several more LLM platforms and millions of agents swarming around the internet, working on behalf of customers and companies and even defence establishments?
An Anthropic employee, Jacob Coxon, recently put out the following post on X: “I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic and neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives…”
He went on to suggest that AI could just about develop the capability to start killing humans, and that fear is not too unreal since AI is definitely being used for military purposes, and the possibility of some of that tech going rogue is high. And what is the chance that China, which is always very secretive about its technological experiments, will not deliberately create rogue agents to do damage to other countries’ economic interests when it has weaponised everything? Also, unlike Anthropic or OpenAI, China may not even choose to disclose what may be going wrong with its AI programmes.
In an earlier article, this writer talked about how countries now see technology and AI as critical to their own national security and global hegemonic power, and will thus not rein in their tech powerhouses because this would mean sacrificing their advantages. At best, they may prevent the Googles, Metas, Nvidias (and in China, the Alibabas, Tencents and DeepSeeks) from harming their own people. When it comes to the rest of the world, they will do little, though they will pretend to care about growing tech power and its impact on the rest of the world.
In defence, AI is already being used extensively to manage battlefield challenges, and as this trend accelerates, most battles may be decided with the use of non-human resources. This may save human lives for the aggressor, but it will cripple those without the power to resist. Iran, despite our concerns about authoritarian cleric rule, did the right thing by building its own deterrents to the overwhelming tech power of the US and Israel. The answer to high-tech appears to be a massive buildup of low-tech swarms of drones and missiles, and the threat to use them even on non-combatants in the region. In short, regional blackmail will be the fallout from the use of high-tech military power by one side.
Even as this AI problem plays out, there are simply too many digital and dark web platforms that offer crime-as-a-service (CaaS), which can escalate the threats to ordinary people. AI agents will be used not just to tackle corporate and personal automation needs, but to amplify crime. Maybe the US, Europe and China have some kind of protective shield for their people, but India and Africa do not. The daily flood of digital scams tells us the real scale of the problem. We digitised on a mass scale without protective cover or building state capacity to deal with the threats.
We are focused on the wrong threats to nations and people. It would make little difference to the world if Iran, Israel, or even Saudi Arabia acquired nuclear weapons, for they are unlikely to actually use them unless it is a question of national survival. But unchecked AI and tech power is a creeping economic, social and political menace larger than nukes. Everyone is joining the AI rat race without worrying about the consequences. Unchecked AI growth is not only damaging job prospects but also leading many profitable AI companies themselves into massive debt, which could send global stock markets crashing when the bills come due and AI does not deliver on revenues.
Since 1945, when two nuclear bombs were used against the Japanese to end the Second World War, six more countries have acquired nuclear weapons (the United Kingdom, France, Soviet Union, India, Pakistan, and North Korea), and one more (Israel) probably has them without saying so. But no nuclear catastrophe has occurred outside the area of civilian nuclear facilities. Unchecked AI has the power to do more damage, not just in the military sphere, but in the civilian space too. In the near term, AI will devastate millions of middle-skill jobs, and entry-level careers will disappear more quickly than the AI-powered jobs needed to replace them.
AI evangelists who regularly promise us that the technology will create more jobs can never be held to account for leading us to AI armageddon even if the opposite happens and superintelligence starts chewing up jobs or renders whole swathes of human intelligence irrelevant to work.
There is, thus, a very strong case for a global regulatory effort in AI that is even stronger than the nuclear proliferation treaty. The chances of that happening may be bleak in the current scenario of distrust of global institutions, which have been misused by the big powers to maintain their hegemony. Countries like India, thus, have no option but to emphasise sovereign tech and control, even if complete sovereignty is an impossibility in a still interdependent world. And, yes, we must start taking data privacy and citizen protection from cyber risks much more seriously.
The writer is a senior journalist
