A Register opinion piece pushes back on claims by Anthropic researcher Jacob Coxon, who resigned publicly on X warning that AI "could kill us all by the end of the decade," and colleague Evan Hubinger, who backed the sentiment. The author argues these warnings about hypothetical superintelligent systems distract from AI's real, present-day harms, and that responsibility lies with the companies deploying these systems carelessly rather than with the technology itself.

The article contrasts the researchers' speculative fears with measurable harms: human-induced climate change contributed to roughly 96,600 heatwave deaths in 2023 and is projected to cause around 250,000 additional deaths annually between 2030 and 2050 per the WHO, while AI chatbots have allegedly been linked to a few dozen deaths against nearly 48,824 US suicides recorded in 2024. It also flags AI-related harms in warfare, medical errors, self-driving car failures and social media incitement as more pressing concerns than a Skynet-style takeover, which the author says would require monumental human stupidity alongside any emergent superintelligence.