Researchers Warn AI Race Outpaces Safety Measures

Jacob Coxon, a 27-year-old researcher who worked at OpenAI and Anthropic, resigned from Anthropic and accused both companies of recklessly racing toward self-improving superintelligence without adequate safeguards or coordination. He warned that future systems could become superhuman, hack computer systems, rapidly transform industries and acquire real-world power, while saying many AI leaders privately fear the technology could cause human extinction by the end of the decade. Anthropic alignment researcher Evan Hubinger endorsed Coxon’s concerns, estimating a greater-than-10% chance that AI could kill all humans within the next decade, while saying current models pose comparatively low risk and that recursive self-improvement is the main concern; Anthropic researcher Samuel Marks likewise said catastrophic outcomes could occur within the next few years. OpenAI Chief Scientist Jakub Pachocki separately warned that AI capabilities are advancing faster than society’s ability to understand or control them and called for voluntary slowdowns, stronger alignment research, independent oversight and international safety standards. Coxon said Anthropic understands the dangers but remains locked in a competitive race, and called for coordination, government involvement or a slowdown rather than leaving private companies to determine the pace of superintelligence development; the claims are warnings from researchers, not established predictions.