Presentation Information
[8p-E310-5]Data Security for Fab Operations in the AI Era
〇Yusuke Maruyama1,2 (1.EAGLYS Inc, 2.Waseda University, Research Organization for Nano & Life Innovation)
Keywords:
Secure Computing,AI Agent,Security
Generative AI, large language models (LLMs), AI agents, and cross-company data collaboration are rapidly advancing fab operations—maintenance, yield, and materials development. Yet capturing this value requires disclosing data and knowledge that have traditionally been isolated and kept confidential: to AI agents, and to external partners. This forces fabs to dismantle the very boundaries on which their security has relied—the isolated internal network, and the corporate wall that keeps sensitive data in-house. The "openness" that creates value thus collides head-on with the "isolation" that has ensured safety. This talk reframes that collision as a shift away from boundary-based (perimeter) defense toward a posture that trusts no one and exposes no data: zero trust for AI agents, and secure computation for data. Centered on secure computation—which allows data to be used while remaining confidential—the talk discusses, through concrete applications in the fab, a foundation for data-driven fab operations that supports the realization of next-generation devices.