Abstract
This paper advances AI safety research in management by integrating Critical Realism, Path Dependence Theory, Endogenous Growth Theory, and the Economics of AI within the framework of Experiential Matrix Theory (EMT). Existing AI governance, grounded in empirical positivism, overlooks multi-layered causal structures, leading to systemic misalignment risks. EMT reframes AI safety by aligning technological progress with evolving human experiential needs, rather than GDP maximisation. Through Path Dependence Theory, we highlight technological lock-in risks, where entrenched AI architectures become difficult to reverse. This underscores the need for preemptive interventions, including modular AI design, competitive AI ecosystems, and dynamic governance models. AI should foster human-AI symbiosis rather than full automation, ensuring resilience and adaptability. We argue that AI safety must incorporate meta-theoretical critique, overcoming historical epistemological lock-ins to align AI with human well-being and economic sustainability. Aligning AI with the Experiential Matrix offers a pathway toward ethical AI governance and sustainable innovation.