Global AI Risks Demand International Safeguards, Experts Warn

Global experts urge coordinated international action to manage AI risks as catastrophic events may unfold gradually rather than dramatically. Dr Simon Nieder, a researcher based in Chesterfield, Derbyshire, argues that the most severe artificial intelligence threats might not appear with a sudden, cinematic explosion like a mushroom cloud. Instead, they could emerge through a series of incremental decisions—such as AI autonomously designing pathogens, exploiting critical infrastructure vulnerabilities, or enhancing weapons systems—where each step seems reasonable until cumulative harm becomes irreversible.

Nieder emphasizes that unlike historical technological disasters, many AI risks may not stem from rogue machines but from systems operating exactly as programmed while humans gradually relinquish control. He calls for governments to establish clear boundaries: some applications, like autonomous weapons and biotechnology, should never operate without human oversight. International agreements could begin by mandating strict human authority for high-stakes decisions, maintaining transparent records of authorization, and sharing data on AI failures and near-misses across borders. Such measures aim to prevent slow-burn catastrophes that escalate undetected.

The historical context of AI governance was set decades ago. Professor Jonathan Michie of Emmanuel College, Cambridge, recalls that as early as 1972, the Serbelloni Group—led by his father, AI pioneer Donald Michie—warned about the societal dangers of artificial intelligence. Michie notes that AI is not a new field; its conceptual roots trace back to Frank Rosenblatt’s perceptron in 1957, the Dartmouth workshop in 1956, and Alan Turing’s foundational work. Yet the rapid evolution of modern computing infrastructure has accelerated AI development beyond earlier projections.

A pivotal moment in UK AI history came with the 1973 Lighthill Report, which criticized AI research and triggered funding cuts, leading to an “AI winter.” Donald Michie, who had worked alongside Turing at Bletchley Park, publicly challenged the report in a televised debate, famously quipping, “Let there be Lighthill. And there was darkness.” His warning resonates today in debates about balancing innovation with safety. Michie advocates for renewed interdisciplinary dialogue—combining computer science and humanities—to assess AI’s long-term societal impacts.

David Kyler, writing from Saint Simons Island, Georgia, highlights an ongoing international initiative to develop a global AI treaty, launched at the UN Global Dialogue on AI Governance in Geneva. UN Secretary-General António Guterres has stressed the need for human control over AI, particularly as it is deployed in warfare, surveillance, and decision-making systems that could reinforce bias or undermine democracy.

Kyler warns that fragmented national or corporate approaches will delay necessary safeguards and increase systemic risks. He points out that the rapid expansion of energy-intensive data centers—driven by competing economic and security interests—complicates efforts to regulate AI effectively. While public opposition to data centers grows due to environmental concerns, Kyler argues this focus distracts from the larger imperative: establishing robust, globally coordinated AI governance before preventable disasters occur.

Like climate change, AI risks transcend borders, requiring multilateral frameworks that preempt misuse and unintended consequences. Without coordinated action, the world risks sleepwalking into preventable crises driven by technology that moves faster than regulation.

#AIgovernance #artificialintelligence #globalregulation #UN #techrisks #AIethics #digitalfuture #Turing