Google pitches proactive AI cyber defense: hunt threats before the attack, not clean up after

DeepMind described a proactive cyber defense model for governments and large enterprises: AI systems continuously scan infrastructure, look for anomalies in user and service behavior, and find vulnerabilities before attackers do. The logic shifts from "respond to the incident" to "prevent the incident."

The urgency is obvious: attackers already use AI to hunt holes and generate phishing, and manual defense against automated offense is mathematically doomed. The symmetric answer is automated defense that runs around the clock and doesn't tire.

For a mid-sized business in Kazakhstan the full arsenal is overkill, but the principle itself — "the machine continuously hunts deviations, the human reviews the findings" — scales to any size. And not only in security: an anomalously large payment, an atypical discount, a strange spike in returns — all are the same deviation-hunting task. In 1Cifer it is handled by auto-checks: the agent constantly monitors operations across company data and raises a flag when something breaks the pattern — without waiting for it to surface in a quarterly report.