Multimodal prompt injection: attacks in images, audio, and video
How attackers bypass text-based guardrails by embedding malicious instructions in images and audio, and the layered defenses required to counter them.
Memory poisoning in AI agents: exploits that wait
How attackers plant instructions targeting agentic AI systems today that execute weeks later, and the defense architecture that stops them.
From LLM to agentic AI: prompt injection got worse
How the shift from single-model LLM integrations to agentic AI systems amplifies prompt injection into a multi-step attack chain.