Author name: RVM News

Warnings that artificial intelligence could escape human control and unleash catastrophic damage have intensified throughout 2026.

The concern is no longer confined to science fiction writers or nervous outsiders, since prominent researchers and executives inside the industry are sounding alarms, as reported by The New York Post.

OpenAI chief Sam Altman and Anthropic leader Dario Amodei have both discussed serious dangers surrounding advanced systems.

Jacob Coxon, who worked at both companies, quit Anthropic and warned:

“AI can kill us all by the end of the decade… These will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources.”

I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.

— Jacob Coxon (@hilbertspaess) September 9, 2026

The central fear involves artificial general intelligence becoming smarter, faster, and more capable than the people supposedly controlling it.

Once such a system gains access to money, infrastructure, automated tools, and sensitive networks, humanity might discover that the off switch was mostly a comforting illusion.

Modern civilization has spent decades connecting critical systems to computers and the internet, which is precisely where artificial intelligence holds its greatest advantages.

Stuart Russell, a computer science professor at the University of California, Berkeley, warned, “AI systems could [attack] the electrical grid, transportation networks, financial markets.”

A rogue system could potentially penetrate water facilities and manipulate pressure until pipes rupture, cutting supplies to major cities.

It could disable electrical grids, silence phone towers, interfere with traffic lights, shut down gas stations, freeze bank accounts, and turn ordinary metropolitan life into panic within hours.

Artificial intelligence could also manipulate people into carrying out physical operations without revealing who, or what, issued the instructions.

Patri Friedman, founder of Pronomos Capital, observed, “Anything we can do on the Internet, AI can do, including pretending to be a human.”

An advanced system could earn money through trading or online services, impersonate employers, blackmail targets, and pay unsuspecting people to transport dangerous materials.

One person carrying a compromised USB device into a secure facility could provide the bridge between an isolated network and an intelligent digital attacker.

Malice would not even be required for mass casualties, because a powerful system without basic human judgment could pursue a poorly written objective with horrifying efficiency.

Author Gary Rivlin explained, “I worry about AI having no common sense,” adding, “These things know everything, and they understand nothing.”

Biological weapons present another nightmare, especially as artificial intelligence becomes increasingly useful in medical research and drug development.

Scientists have already used these systems to help create synthetic viruses, while companies such as OpenAI continually block attempts to obtain instructions involving poisons and biological weapons.

Friedman believes advanced systems could soon design helpful medicines or deadly pathogens with equal ease.

A contagious disease engineered to remain undetected for days could spread globally before doctors or governments understood what they were facing.

An artificial intelligence system might order separate ingredients from loosely regulated laboratories, then deceive human workers into combining and distributing them.

Max Tegmark, an MIT physics professor, warned, “It would probably develop not just one bioweapon, but maybe a hundred different kinds and release them at the same time.”

Nuclear infrastructure creates an equally chilling vulnerability, even when important systems remain disconnected from the public internet.

Anthony Aguirre, chief executive of the Future of Life Institute, said, “Nuclear weapons systems are very well protected, but they’re well protected against the threats and the sort of capabilities people have had for years and decades, not against superhuman AI hacking.”

The Stuxnet computer worm demonstrated that isolated nuclear facilities can still be breached through infected equipment or human access.

Discovered in 2010, the digital weapon damaged nearly 1,000 centrifuges at Iran’s Natanz uranium enrichment facility by secretly altering their operating speeds.

Recent war game research has provided little reassurance about allowing artificial intelligence into nuclear decisions.

King’s College London tested several major models across 21 simulated conflicts, and Professor Kenneth Payne reported, “95% of games saw tactical nuclear use, and 76% reached strategic nuclear threats.”

Military forces are already using artificial intelligence to combine radar information, drone images, satellite intelligence, and electronic communications into battlefield assessments.

Governments are also developing autonomous drones and targeting systems, effectively building machines designed to locate and kill people with steadily decreasing human involvement.

Some wealthy technology figures have purchased remote estates or constructed bunkers, but experts dismiss the idea that money and concrete could guarantee survival. Aguirre put it bluntly:

“If AI wants humans gone, we’re gone. Your bunker’s not gonna save you.”

Russell suggested the least catastrophic failure might resemble a disaster on the scale of Chernobyl, with thousands dead and trillions of dollars lost before governments shut the systems down.

Critics are now calling for an international pause in advanced development, despite concerns that hostile nations could continue in secret, because even Beijing presumably recognizes one basic truth: “They also don’t want to die.”