Kadena Air Base, Japan -- Artificial intelligence is changing the way the world operates: and the way it fights. As technology advances faster than ever before, the challenge isn’t just creating smarter systems, but protecting them from being turned against their users.
That’s exactly what the 18th Communications Squadron is preparing for with “Ghost in the Model,” a weeklong case study held at Kadena Air Base, Japan, Oct. 20–24 as part of Cybersecurity Awareness Month.
“Information is ammunition, and we cannot afford for ours to be compromised,” said Lt. Col. Henry Sims, 18th CS commander. “We’re preparing our Airmen for the fight of the future — one where the enemy may already be inside the machine.”
The exercise centers around MITRE ATT&CK — short for Adversarial Tactics, Techniques, and Common Knowledge — a framework that helps cybersecurity professionals model and counter real-world attacks. The “Ghost in the Model” team used this system to develop countermeasures for four simulated scenarios where adversaries might corrupt or manipulate artificial intelligence.
Some of the threats explored included poisoning AI training data to misidentify friendly systems as hostile, embedding hidden commands within legitimate datasets, or deliberately disrupting models to produce false conclusions.
“The systems could be fifty years old or two weeks old,” said Master Sgt. Amy Kern, 18th CS cyber security section chief. “We’re identifying the best strategies for any situation, even the ones we haven’t seen yet. The biggest cyber threat is the one you don’t see coming.”
By testing how AI tools react when compromised, the team is developing tactics that will protect networks and missions across the Indo-Pacific and beyond. The goal isn’t just to respond to cyberattacks: it’s to predict and prevent them before they happen.