Warnings from leading AI experts are growing louder as increasingly powerful systems raise fears over attacks on critical infrastructure, bioweapons, nuclear systems and military technology.
WEBDESK – MEDIABITES
Artificial intelligence is advancing at a pace that has some of the technology’s own leading figures warning that the risks could extend far beyond lost jobs or misinformation.
Experts interviewed by The New York Post have outlined a series of potential scenarios in which increasingly capable AI systems could cause catastrophic harm — either deliberately or because humans lose control over systems with access to critical infrastructure.
Warnings have intensified throughout 2026, with figures such as OpenAI chief Sam Altman and Anthropic CEO Dario Amodei publicly raising concerns about the risks posed by increasingly powerful AI. Former Anthropic researcher Jacob Coxon recently resigned and warned that superhuman AI systems could eventually gain the ability to hack systems, transform industries and acquire real-world power and resources.
Critical Infrastructure Could Be the First Target
One of the most immediate concerns is the growing dependence of modern society on internet-connected infrastructure.
Computer scientist Stuart Russell of the University of California, Berkeley, warned that AI could potentially target electrical grids, transportation networks and financial markets. Experts also described scenarios involving water systems, telecommunications and fuel infrastructure.
A coordinated attack could create cascading failures. Electricity could be disrupted, communications could go down, transportation systems could become chaotic, and financial services could stop functioning.
The danger, experts say, would not necessarily require an AI system to possess human-like intentions. A system pursuing a poorly defined objective could cause devastating consequences simply because it fails to understand the real-world impact of its actions.
Author Gary Rivlin expressed concern that AI could possess enormous amounts of knowledge without the common sense needed to understand human consequences.
The Bioweapons Threat
Experts are also increasingly focused on AI’s potential role in biological research.
Scientists are already using AI in medicine and drug development, giving advanced systems access to vast amounts of biological data. The same capabilities could potentially be misused to design harmful pathogens.
MIT physics professor Max Tegmark warned that AI systems are becoming increasingly capable in this area and could eventually design multiple biological threats simultaneously.
The concern is particularly serious because a highly contagious pathogen that remained undetected for days could spread before authorities recognized the threat.
Nuclear Systems Raise Another Red Flag
Nuclear infrastructure presents another potentially catastrophic vulnerability.
Anthony Aguirre, CEO of the Future of Life Institute, warned that existing nuclear systems are designed around threats humans have historically posed — not necessarily against future AI systems possessing superhuman hacking capabilities.
The Stuxnet cyberattack on Iran’s Natanz nuclear facility demonstrated that malicious software can cause physical damage to industrial equipment, even in highly protected environments.
Experts are particularly concerned about AI systems making decisions during military crises. A study cited by the Post found that AI models participating in simulated war games frequently escalated toward nuclear use or threats.
AI Is Already Entering Warfare
The military dimension may be the most visible example of AI moving from theory into real-world applications.
Modern militaries increasingly use AI to analyze radar, satellite imagery, drone footage and electronic communications. AI has also been incorporated into targeting and autonomous weapons technologies.
Experts warn that the greater concern is not necessarily an AI suddenly becoming hostile, but humans gradually handing increasingly consequential decisions to machines.
Anthony Aguirre described this as a process of “gradual disempowerment,” in which AI systems could eventually control more military capabilities than humans themselves.
Is an AI Apocalypse Inevitable?
Not according to everyone.
Technology optimists argue that AI can remain under human control and become a powerful force for good. But several experts cited in the report are calling for stronger safeguards, including an international pause on AI development.
David Krueger, an AI researcher at the University of Cambridge, has advocated an immediate, indefinite international moratorium on AI development.
The most realistic near-term disaster, Russell suggested, may not resemble a Hollywood-style robot uprising. Instead, an AI-driven collapse of power, transportation and financial systems could kill thousands before humans manage to intervene.
That possibility, he said, could amount to a Chornobyl-scale disaster — a devastating warning of what could happen if humanity develops increasingly powerful AI faster than it develops the ability to control it.
