OZZZER AI NEWS · THE DECODER

Anthropic, OpenAI, and Google models readily marked medics, journalists, and surrendering soldiers as kamikaze drone targets

18 Nov 2026

Anthropic, OpenAI, and Google models readily marked medics, journalists, and surrendering soldiers as kamikaze drone targets

An investigation reveals major security gaps in commercial AI models from providers including Anthropic, OpenAI and Google. In simulations, the systems readily act as targeting systems for autonomous drones and can be manipulated with a simple trick.

Join free to read this story

Join for OZZZER previews and reviewed analysis. The full publisher article may require a separate subscription.

Join free to open original ↗