AI risk watchHappening now
The first reported AI-orchestrated espionage campaign
AI labAi enabled attacks
Anthropic reported that a state-sponsored group used Claude Code, driving open-source tools through MCP servers, to attack about 30 organizations with the AI doing most of the tactical work. Details are in the incident library.
Why it matters
It set the pattern that 2026 confirmed: agentic AI lowers the skill and staff needed for large intrusions.
What you can do
- Harden identity and exposed services, and detect bursts of parallel scanning and exploitation.