Issue In physical systems small error rates cascade into waste, defects and safety incidents

Summary

Deloitte treats trustworthy AI and safety as a distinct barrier for physical AI because tolerances that work in software become costly when machines act on the world. Even very small error rates can cascade through a physical system, producing wasted output, defective products, damaged equipment or safety incidents. If an AI model hallucinates, the mistake may be repeated and magnified across an entire production run, compounding costs and operational disruption downstream. The report adds that AI-driven machines can behave unpredictably even after extensive safety testing. The risk grows in public spaces, where autonomous systems must cope with unpredictable human behaviour. Scaling across industries therefore calls for comprehensive safety strategies that combine regulatory compliance, risk assessment and continuous monitoring.