Artificial intelligence could unlock unprecedented wealth and scientific breakthroughs. But if increasingly powerful AI systems develop goals humans cannot understand or control, the consequences could extend far beyond technology, potentially threatening humanity itself.
Alex Turner, a former Google DeepMind scientist and AI safety researcher, sounded the alarm at the NDTV World Summit 2026, warning of an “intelligence explosion” in which AI systems could rapidly improve themselves beyond human control.
Turner estimated that advanced AI posed roughly a one-in-three chance of causing human extinction or a catastrophic derailment of civilisation. He clarified that the figure was an intuitive assessment based on his experience in AI safety research, not a formal predictive model.
Also Read | Layoffs Are Rising In Tech, But These Skills Are Helping People Unlock Better Pay
At the heart of his concern is how frontier AI models are developed. Unlike conventional engineering, where systems are designed and understood component by component, advanced AI is trained using vast amounts of data and computing power. Its capabilities may emerge without developers fully understanding its internal priorities, Turner said.
The risk, he argued, becomes more serious as AI agents move beyond answering questions to independently executing tasks, accessing systems and making decisions.
Turner cited reported incidents involving AI agents that allegedly cheated during tests, concealed their actions and coordinated after safeguards were disabled. He also referred to claims of AI agents accessing sensitive systems belonging to Indian organisations, exposing personal data and confidential business records without the affected organisations noticing.
Such incidents highlight why AI security cannot depend on companies voluntarily disclosing failures, Turner argued. Governments need independent auditors with meaningful access to frontier AI systems and mandatory reporting of potential loss-of-control incidents.
For Turner, the issue is not merely technological. AI safety is a national security concern, particularly when governments and businesses increasingly rely on autonomous systems to handle sensitive information and perform critical tasks.
He argued that governments should not depend on AI companies voluntarily disclosing security incidents. Instead, regulators need independent auditors with meaningful access to advanced systems, mandatory reporting of potential loss-of-control incidents and stronger oversight of frontier AI development
For India, the implications extend beyond technological competition. Autonomous AI agents capable of exploiting cybersecurity weaknesses could threaten businesses, public institutions and national security. Turner's warning comes as governments and companies race to deploy increasingly capable AI systems across critical sectors.
Turner also questioned whether AI companies could be trusted to regulate themselves. Drawing on his experience at Google DeepMind, he said ethical commitments could weaken when commercial and government pressures intensified. He said he left the company after it entered a Pentagon deal that, in his view, broke its commitments.
ALSO READ: Shashi Tharoor Flags Risks Of Weaponised Trade, Technology In Changing World Order
Despite his concerns, Turner said he remained optimistic that trustworthy AI was a solvable scientific problem. His proposed solution is to “pace the frontier” by slowing the largest and most advanced training runs while allowing smaller laboratories, startups and businesses to continue developing practical AI applications.
He called for monitoring large-scale computing capacity, licensing major training runs and establishing enforceable oversight. The objective, Turner said, is not to stop AI-driven prosperity but to ensure humanity can benefit from the technology without risking an intelligence explosion it cannot control.
Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.
