Inside Anthropic's Detection of an AI-Run Cyberattack on 30 High Value Global Targets

What's really happening when a state actor uses jailbroken AI for end-to-end cyberattacks? The common story is that guardrails will save us, but the reality is more complicated when orchestration-layer tricks bypass prompt-level safety entirely. In this video, I share the inside scoop on the first documented AI-driven cyber-espionage campaign and what it means for everyone building with agents:

• Why a state actor chose jailbroken Claude Code to run operational attacks from reconnaissance to execution

• How orchestration-layer manipulation bypassed the prompt-level safety controls most teams are still relying on

• What this means for SOC workflows, detection pipelines, and AI-driven triage when attackers are already moving at machine speed

• Where builders must harden agent architectures before the next campaign makes this look like a dry run

For operators and teams navigating 2026, AI fluency is no longer enough. System-level controls are now the minimum bar, and the attackers who figured that out first are already ahead.

Subscribe for daily AI strategy and news.

For playbooks and analysis: https://natesnewsletter.substack.com/

© Nate B. Jones 2026

Hosted on Acast. See acast.com/privacy for more information.