Autonomous agents rely on machine-actionable data to drive workflows, and semantic metadata has played a crucial role in enabling data discovery and interoperability. The use of schema.org and other semantic metadata standards has been instrumental in implementing the FAIR principles, making data more findable, accessible, and reusable. However, the emergence of Large Language Models (LLMs) has raised questions about the continued need for semantic metadata in agentic data retrieval1. As LLMs become increasingly capable of navigating unstructured data, the role of semantic metadata may shift, potentially altering the security implications and risk surfaces associated with data-driven workflows. The development of LLMs, such as those by Google, is reshaping the capability and risk landscape, and security implications are a critical consideration. The evolution of LLMs and their impact on semantic metadata highlights the need for practitioners to reassess their data management strategies to ensure the security and integrity of their workflows.