Research reports on the agentic threat landscape.

Emergent attack patterns against autonomous agents, including multi-turn jailbreaks, tool-abuse chains, and privilege-escalation trajectories.

Deployment findings from the semantic inversion preprocessing layer: latency, false-positive rates, and polite-language attack mitigation.

Experimental evidence of steganographic coordination in frontier multi-agent systems and the controls that detect it.