The Rise of the Non-Human Insider: When AI Agents Become the Threat
KnowBe4, Thursday, August 20th, 2026
KnowBe4 argues machine identities now outnumber humans 82 to 1, making authenticated AI agents a serious insider risk.
KnowBe4's Erich Kron argues that cybersecurity's familiar villain, the external attacker breaking in, no longer describes where the risk sits.
Machine identities now outnumber human ones by 82 to 1 in typical environments. That means the next breach is more likely to originate from an authenticated AI agent operating within its granted permissions than from an intruder exploiting an unpatched vulnerability.
The post introduces the non-human insider as a category requiring its own controls. It argues identity governance has to extend to agents with the same rigor applied to employees.