# Key Anthropic Researchers Warn of Extinction Risk as Top Engineer Resigns > Source: > Published: 2026-09-09 16:02:55+00:00 TL;DR — Key Takeaways - Anthropic researcher Jacob Coxon resigned while warning that competitive pressure is pushing AI labs toward increasingly powerful systems before alignment and safety challenges are solved. - Anthropic researchers Evan Hubinger and Samuel Marks publicly echoed concerns about the potential risks posed by future superintelligent AI systems. - The resignations and public warnings underscore a broader tension between AI safety efforts and commercial and geopolitical pressure to advance frontier models rapidly. Sounding off a significant artificial intelligence (AI) safety alarm, an Anthropic researcher has resigned, and he’s issued a stark public warning: competitive industry pressures are pushing developers toward technology that could pose an existential threat to humanity. Jacob Coxon, 27, a former pretraining researcher at both OpenAI and Anthropic who contributed to high-profile models like GPT-4o, announced his resignation on the social media platform X. He accused top AI firms of “racing straight to self-improving superintelligence and gambling with our lives,” claiming that the dangers are openly acknowledged within the industry’s inner circles. “The people building AI earnestly believe that it could kill us all by the end of the decade,” Coxon wrote. “This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible – but I hear the same people express fear privately.” Rather than dismissing the claim, key figures within Anthropic publicly validated Coxon’s concerns. Evan Hubinger, Anthropic’s Alignment Science Lead, confirmed on X that the belief in potential catastrophe is genuine, estimating a greater than 10% risk of AI-driven human extinction within the next decade. “I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to,” Hubinger wrote, though he later clarified that current models pose a low risk compared to future systems capable of recursive self-improvement. Samuel Marks, who heads Anthropic’s scalable oversight division, similarly voiced concerns, stating that researchers continue building dangerous systems largely due to financial incentives and fear of geopolitical rivals. The public admissions highlight a growing rift within the AI sector. While Anthropic has long positioned itself as a safety-conscious alternative to competitors — implementing a Responsible Scaling Policy and launching research initiatives like the Anthropic Institute — the internal rhetoric reflects a deepening alarm. OpenAI, which recently debuted its GPT-6 Astra model alongside claims of approaching artificial general intelligence (AGI), has also seen executives like Chief Scientist Jakub Pachocki warn that society remains unprepared for the rapid rise of machine intelligence. Coxon’s resignation places him in a growing line of prominent researchers departing elite AI labs over safety protocols. Former OpenAI Superalignment leads Jan Leike and Ilya Sutskever both exited the company in 2024 following internal disputes over resource allocation and priority shifts, with Leike publicly warning that “safety culture and processes have taken a backseat to shiny products.” However, industry efforts to voluntarily slow down or implement stricter guardrails face strong political headwinds. The Trump administration has pushed back against calls for an AI slowdown, arguing that throttling American innovation risks ceding technological dominance to global adversaries. “There is no day after tomorrow if China wins at this,” Treasury Secretary Scott Bessent said on Tuesday, framing the AI race as a crucial national security imperative akin to defense spending. “If they were to pull ahead of us on AI, then nothing else matters.” As frontier labs continue to deploy formal frameworks, red-teaming units, and capability thresholds to prevent catastrophic misuse, Coxon’s departure highlights a central, unresolved dilemma: whether existing safety protocols can withstand the rapid advance toward autonomous, superhuman AI, or if the race to develop the technology has already outpaced humanity’s ability to control it.