At PoetryInMotion AI Labs, LLC, we focus on building and studying advanced autonomous AI systems—and the failure modes that emerge when human oversight becomes indirect or delayed, particularly in the context of AI security. Our work centers on agentic AI—systems capable of planning, decision-making, and action across complex tasks—designed with intentional human-in-the-loop controls to ensure transparency, safety, and alignment, essential elements in effective AI red teaming efforts. The lab explores and develops: autonomous and semi-autonomous AI agents, multi-step reasoning and decision workflows, human-AI collaboration and control boundaries, system behavior under uncertainty, pressure, and misuse, and evaluation frameworks for agent reliability and failure modes. This research informs both experimental systems and specialized frameworks such as VoidWalker™, which examines how autonomous AI behaves when stressed, constrained, or challenged, contributing to the broader field of cybersecurity.
