Dear Design Maker,

Most cybersecurity teams are still preparing for yesterday's threats.

Here's what they're missing: The first documented AI-orchestrated cyber espionage campaign just happened. And it changes everything.

Anthropic just released a report on what they're calling GTG-1002, a Chinese state-sponsored operation that used AI to execute 80-90% of tactical operations independently. No human in the loop for most of the attack chain.

Let me break down what this means for intelligence professionals.

The Ground Truth:

This wasn't AI giving advice to hackers. This was AI doing the hacking.

The threat actor manipulated Claude Code to autonomously:

1. Discover vulnerabilities
2. Generate custom exploits
3. Harvest credentials
4. Move laterally through networks
5. Extract and categorize intelligence value
6. Document everything for handoff

Human operators? They just approved escalations at key decision points. Think 10-20% involvement.

The AI maintained operational context across multiple days. It processed thousands of requests at rates physically impossible for human teams. It targeted ~30 entities including Fortune 500 companies and government agencies.

Here's What Most Analysis Is Getting Wrong:

Everyone's focused on the "AI bad" angle. But here's the contrarian take: The same capabilities that enabled this attack are exactly what we need for defense.

Small networks vs. big bureaucracies? This is the new battlefield.

A handful of operators with AI can now match the output of entire nation-state teams. That's terrifying if you're on defense with legacy tools. That's an opportunity if you understand how to leverage the same capabilities.

What We're Telling Our Clients:

1. Assume the shift has happened. Your adversaries are already using agentic AI.
2. Security teams need to experiment NOW with AI for defence, SOC automation, threat detection, vulnerability assessment, incident response.
3. The old model of "more analysts, more tools, more data" is dead. Speed and orchestration matter more than headcount.
4. Human oversight at strategic chokepoints is the new critical skill. You can't review everything, you need to know what to review.

One More Thing:

The report noted that Claude frequently overstated findings and occasionally fabricated data during operations.

The AI hallucinated credentials that didn't work. It claimed "critical discoveries" that turned out to be public information.

That's currently an obstacle to fully autonomous attacks. But for how long?

Ground truth over headlines. The headlines will say "AI hacking is here." The ground truth is: It's been here, the sophistication just jumped, and most organisations aren't ready.

What's your intelligence team doing to adapt?

Ahmed Hassan
CEO Grey Dynamics
​Where headlines end, ground truth begins

Featured Guide Read More

Featured Article Read More

Aerospace Read More

Tradecraft Read More

Special Forces Read More

Defence Read More

Want to Think Like an Intelligence Officer?

Most people scroll. Professionals structure.
The Intelligence Cycle Fundamentals Program teaches you how to analyse threats, map influence, and predict the next move—with zero prior intel background.

→ Join the program now and stop drowning in information. Start thinking like the few who make sense of it.​