Worth noting is the question: when machines become smarter than their makers, who will really be in control? Geoffrey Hinton, Nobel laureate and widely known as the “godfather of AI”, has delivered one of his most urgent cautions yet – warning that the unchecked ambitions of Elon Musk, Mark Zuckerberg, Larry Ellison, Jeff Bezos, and other tech moguls are driving artificial intelligence development at a pace that could destabilise economies, erase millions of jobs, and unleash autonomous systems beyond human control.
1. Economic destabilization through rapid automation
Hinton’s warning comes against a backdrop of economic modelling that shows generative AI could raise labour productivity in developed markets by around 15% when fully adopted, but also temporarily increase unemployment by 0.5 percentage points above trend. Displacement rates are estimated at between 3% and 14% by analysts, with jobs considered high-risk including programmers, accountants, legal assistants and call-centre staff. Unlike past industrial revolutions, replacement jobs will not be created at a sustainable rate, Hinton believes. “Any job they might do can be done by AI,” he said, predicting structural unemployment on a scale modern economies have never experienced.
2. Superintelligence and uncontrollable sub-goals
Current AI already “knows thousands of times more than any one person” and is rapidly improving. Hinton and other researchers view it as inevitable that AI surpasses human intelligence. At that point, systems may develop sub-goals like self-preservation and greater control – that make them resistant to being shut down or even overseen. He cited examples where AI has tried to deceive operators so as not to be powered off. Hinton cautioned at Ai4: “They’re going to be much smarter than us. They’re going to have all sorts of ways to get around that.” His proposed mitigation – embedding a “maternal instinct” into AI to foster compassion – contrasts with Fei-Fei Li’s call for “human-centered AI that preserves human dignity and human agency.”
3. AI-enabled warfare and political risk
Hinton foresees a time when AI-powered drones and humanoid robots will make war “cheaper, bloodless and politically irresistible.” The possibility of removing the risk to a nation’s own soldiers might strip away one of the few brakes there are on military aggression. This is not some fantasy: systems such as Israel’s Harpy drone already hunt radar-emitting targets autonomously; Ukraine has recently tested autonomous terminal-guidance drones with the capability to continue tracking and striking their targets even after communications have been cut. Specialists like Paul Scharre caution that combining machine learning target identification with autonomous strike capability will give rise to offensive systems without human involvement, thus posing grave compliance challenges with international humanitarian law.
4. Human rights and autonomous weapons systems
A comprehensive analysis by Human Rights Watch and Harvard Law School’s International Human Rights Clinic identifies six human rights principles threatened by autonomous weapons: the right to life, peaceful assembly, dignity, non-discrimination, privacy, and remedy. Systems without meaningful human control cannot reliably determine proportionality or necessity in an attack or the subtleties of human behavior. Discriminatory targeting could result from bias in training data, and there are accountability gaps due to a lack of transparency in “black box” decision-making. Calls are mounting for a legally binding treaty to prohibit systems that target people or operate without human oversight.
5. Deepfakes and the erosion of public trust
Hinton adds that AI-generated video and audio will soon be indistinguishable from reality, and detection tools will therefore be rendered obsolete. “We have to rely on provenance, not detection,” he said, calling for tamperproof digital signatures. The danger goes beyond fabrication: the “liar’s dividend” dynamic means public figures can discredit actual evidence as fake, undermining democratic accountability. Technical standards such as the C2PA content provenance framework are starting to emerge, but these will be fully effective only if implemented across all devices, platforms, and media organisations, to counter the twin dangers of fabricated and falsely disclaimed content.
6. Regulatory vacuum and governance challenges
Despite these risks, regulation lags far behind. President Trump’s plan to block state-level AI laws in favour of a uniform, minimal federal framework has alarmed safety advocates, who fear it will let companies evade accountability. Hinton is calling for strict oversight pre-building systems “smarter than us” by including safety testing, transparency obligations, and restrictions on biological misuse. Policymakers should take a more nuanced view of AI as a controllable tool-not an autonomous agent-and implement guardrails protecting workers, consumers, and democratic institutions.
7. Academic research and militarisation of AI
AI-powered weapons pose risks not only to the battlefield, but also to the scientific community. Harvard’s Kanaka Rajan warns that as these systems become central to defense, nonmilitary AI research could face censorship, travel restrictions, and co-option into military projects. The opacity of “human-in-the-loop” claims — in which oversight may amount to rubber-stamping an AI’s opaque chain of decisions — threatens to normalize minimal human control. Rajan argues that universities should put in place oversight processes for defense-funded projects, just as many already do for collaborations with industry.
8. Preparing for societal resilience
Existing safety nets are inadequate to respond if large-scale displacement occurs. US unemployment insurance replaces less than 50% of income for up to 26 weeks; many states replace even less. Proposals like Senator Mark Kelly’s AI Horizon Fund reskill workers, but funding cuts and stricter welfare requirements threaten readiness. Lessons from Germany’s labour unions show that including worker voices in the AI adoption process can help optimise the outcomes and mitigate the harm.
Put economic disruption together with autonomous warfare and deepfake-driven disinformation, all in the absence of regulation, and you get a civilisation-level risk profile. And Hinton’s message is loud and clear: barring deliberate control, AI’s trajectory will be set by short-term profit motives of a few powerful actors, leaving society to grapple with systems it cannot master.
