Former Anthropic researcher Jacob Coxon reportedly resigned in September 2026, warning that leading AI laboratories are racing toward self-improving superintelligence while potentially risking humanity's future.
Coxon emphasised that current AI systems do not pose extinction-level threats. His concerns focus on future models capable of independently developing more powerful successors.
Recursive self-improvement could accelerate technological progress dramatically, potentially creating feedback loops that exceed human understanding, supervision and intervention capabilities.
Anthropic researcher Evan Hubinger reportedly estimated a greater than 10% probability of AI causing human extinction within a decade, reflecting his personal assessment rather than scientific consensus.
Intense competition among AI laboratories creates pressure to accelerate innovation, even as researchers acknowledge unresolved challenges involving alignment, accountability and controllability.
Public reactions remain divided. Supporters demand stronger safeguards and international oversight, while critics question whether such warnings exaggerate risks or reflect commercial competition.
The fundamental challenge is ensuring AI safety advances alongside capabilities. Without effective governance, rapid innovation could outpace humanity's ability to maintain meaningful control.
