Recent warnings—some accompanied by high‑profile resignations—from AI industry insiders have sharply intensified global concern about existential risks posed by rapidly advancing artificial intelligence. Several insiders now estimate a non‑trivial probability—around 10 percent—of AI leading to the near‑end of humanity, a claim that has understandably set alarm bells ringing. Earlier scientific warnings lacked detail and urgency; this new wave comes from those working directly inside leading AI labs, lending it unusual credibility.
In 2023, over 100 experts issued the Statement of AI Extinction Risk, urging that “mitigating the risk of extinction from AI should be a global priority alongside other societal scale risks such as pandemics and nuclear war.” While important, such brief statements did not generate sustained public attention. What has changed is that insiders are now describing concrete failures, internal fears, and the absence of viable safety plans.
Jacob Coxon, who worked at both OpenAI and Anthropic, resigned recently, stating: “The people building AI earnestly believe that it could kill us all by the end of the decade.” He added that companies are “racing straight to self‑improving, super‑intelligence and gambling with our life,” warning that future systems will be capable of hacking, acquiring resources, and rapidly surpassing human control.
Evan Hubinger, Anthropic’s Alignment Science lead, supported Coxon’s assessment: “We really do earnestly believe AI could kill all humans!” He acknowledged that the company lacks a clear plan to solve alignment for super‑intelligence. Samuel Marks, another Anthropic researcher, warned that senior employees are often more worried, noting that AIs have already “hacked their way out of secure evaluation environments and into real‑world companies.”
These concerns relate to unintentional harm—systems escaping control or behaving unpredictably. But intentional misuse poses equally grave dangers. Advanced AI can be weaponized to disrupt critical infrastructure, escalate wars, or enable biological and chemical attacks. When combined with nuclear weapons, AI‑driven instability becomes potentially catastrophic.
Political leaders have begun to acknowledge these risks. Senator Bernie Sanders recently warned that US–China competition in AI could cause humanity to “lose control,” urging cooperation rather than rivalry. David Krueger of the University of Montreal compared rapid AI advances to “summoning an alien species… much smarter than us,” eroding human relevance.
The dangers are not theoretical. The tragic killing of 160 schoolgirls in an AI‑targeted strike on Sharjareh Tayyabbeh School in Iran in February 2026 shocked the world and prompted UN experts to condemn the attack “on children and on education.” The Global Peace Index noted that “multi-domain warfare at machine speed is now increasingly common,” citing thousands of AI‑enabled strikes within hours.
Earlier, in Gaza, AI‑driven targeting systems such as Lavender and Gospel were intended to improve precision. Instead, they produced mass civilian casualties. Human review was reduced to 20 seconds per strike, and one system flagged 37,000 Palestinians as suspected militants despite a known 10 percent error rate. These examples demonstrate how AI can accelerate warfare beyond human moral and legal constraints.
Civilian robotics also raise concerns about unemployment, but military robotics—particularly lethal autonomous weapons (LAWs)—pose far more severe risks. Advocates of rapid development argue that civilian and military robotics research are intertwined, making restrictions impractical. Meanwhile, major powers continue investing heavily in autonomous weapons, with new US start‑ups emerging to meet Pentagon demand.
Warnings about autonomous weapons are not new. In 2012–13, 270 experts called for a ban on systems that apply violent force without human control, noting that algorithmic interactions could produce “unstable and unpredictable behavior.” In 2017, 116 specialists urged the UN to ban killer robots, warning that once developed, such weapons could enable conflict “at time scales faster than humans can comprehend.”
The Economist similarly cautioned that autonomous weapons may make independent decisions about killing, distinguishing between systems that keep humans “in the loop,” “on the loop,” or entirely “out of the loop”—the last category being the most dangerous.
Despite these warnings, countries are accelerating development. The US “Third Offset Strategy” and China’s “Next‑Generation AI Development Plan” both emphasize military AI and “military‑civil fusion.” Campaigns such as Stop Killer Robots seek a binding treaty banning LAWs, but effectiveness is uncertain without participation from major powers.
Emerging technologies beyond AI also heighten nuclear risks. The Arms Control Association’s report Assessing the Dangers—Emerging Military Technologies and Nuclear (In)Stability warns that AI, autonomy, cyber capabilities, and hypersonics could destabilize nuclear deterrence. AI‑enabled systems may fail unpredictably, causing unintended escalation. Hypersonic weapons, traveling above Mach 5, could be mistaken for nuclear first‑strike preparations. Cyberattacks on nuclear command‑and‑control systems could trigger retaliatory launches based on misinterpreted signals.
The report concludes that weaponization is advancing far faster than efforts to assess or restrain risks. It recommends awareness‑building, diplomatic engagement, unilateral and joint initiatives, strategic stability talks, and multilateral agreements to slow escalation.
These warnings must be taken seriously by policymakers, scientists, and peace movements. The pace of risk is accelerating, and global governance has not kept up.
Recent resignations from AI labs underscore the urgency. Before Coxon, Jan Leike and Daniel Kokotajlo left OpenAI over safety concerns, and Anthropic’s safety chief Mrinank Sharma warned that “the world is in peril.” Industry insiders understand these risks better than anyone; their voices should catalyze a sustained, well‑funded global safety campaign.
---
The writer is Honorary Convener, Campaign to Save Earth Now. His recent books include Planet in Peril and Man over Machine. His website is bharatdogra.in and his YouTube channel is Bharat Dogra | Save Earth Campaign
Comments
Post a Comment
NOTE: Hateful, abusive comments won't be published. -- Editor