Definition

Highly targeted email (or messaging) attacks that impersonate trusted parties using personal/organizational context — increasingly authored or amplified by generative AI at low cost.

Key Points

  • aegisai M3AAWG study claim: AI-generated share of observed phishing rose ~2.8%→13.9% in 2025; higher inbox reach; many cleared authentication via compromised accounts
  • Rule/signature SEGs struggle when content is novel and from legitimate-looking accounts
  • Defense trend: agentic mail analysis (aegisai) vs. static rules; competes with Abnormal, Proofpoint, Mimecast, Ocean

Sources