ModelRed Blog
Latest news and updates from ModelRed.
Gatekeeping AI Security: Lessons from DEFCON and the Copilot Hack
At DEFCON, Microsoft’s Copilot Studio agents were hijacked with prompt injections, exposing private data. The incident shows why AI security can’t remain locked behind closed doors.
Introducing ModelRed: Making AI Safer for Everyone
Most AI models are deployed without real-world stress testing. ModelRed introduces the ModelRed Score — combining general probes and domain-specific simulations — to give enterprises, researchers, and developers measurable trust.