Organizations investing in AI to strengthen workplace safety may be unintentionally weakening the human capacity that prevents a mistake from becoming a fatality, according to an article in the July/August 2026 issue of Occupational Health & Safety. While AI can improve hazard detection and injury prevention through predictive analytics, computer vision, and machine learning, the article argues that the safety profession is neglecting the "second half of the safety equation": recovery capacity. Recovery capacity is defined as the organizational ability to detect deviation, contain consequences, and extract learning before a mistake becomes a fatality. Prevention capacity improves with better engineering controls, tools, and procedures, while recovery capacity improves through human practice, such as trained supervisors, crews with shared language for raising concerns, and leaders who create conditions for fast, unfiltered bad news. The article warns that heavy reliance on automation could quietly erode recovery capacity, leaving organizations measurably better at preventing common injuries but structurally less capable of containing rare, fatal events.