The model may not be your biggest risk: Systemic factors in AI safety

bigdata · x · 2026-08-19

Ben Lorica argues that the biggest AI risks often lie outside the model itself, citing incidents where architecture, permissions, and data provenance led to major failures. The article discusses how a mediocre model with broad access can be more dangerous than a strong model in a restricted environment, highlighting the importance of system-level governance.

Related event: AI's Biggest Risk Lies in Surrounding Systems, Not Model Power(2 posts)→

Original post →

More from Safety

Safety channel →