ai-security

Security at the intersection of AI systems and adversarial threats covers attacks that exploit model behavior, training data, and inference pipelines. This includes prompt injection, adversarial inputs designed to fool perception systems, and techniques that manipulate AI outputs to deceive downstream humans or automated agents. Practitioners in this space work on threat modeling, red-teaming, and defenses that account for the blurred boundary between digital and physical attack surfaces.

A person in an adversarial-patterned t-shirt walks a surveillance corridor while a thermal camera overlay fails to detect a h

adversarial-ai, prompt-injection, agentic-ai, computer-vision, ai-security

The War Where You Are the Target

From t-shirts that defeat thermal cameras to instructions injected into tool outputs, adversarial AI attacks now run directly through human bodies and input channels, undermining the assumption that humans permanently control the upstream position.