Fujitsu creates secure tech for collaborative AI agents enabling inter-company supply chain optimization without sharing sensitive data.
Fujitsu announced a novel multi-AI agent collaboration technology designed to enable seamless and secure coordination among AI agents from different companies without disclosing confidential information. The approach employs a negotiation-based mechanism where a proposing AI agent approximates other agents’ characteristics through iterative suggestion and answer exchanges to identify globally optimal states for applications such as supply chain management. This is achieved under conditions of incomplete information, preserving competitive data privacy. To maintain security, Fujitsu leverages expertise in large language model guardrails to detect malicious queries and prevent unintended data inference. The system repeatedly simulates AI agent behaviors and updates communication in encrypted, non-inferable formats, ensuring robust protection of sensitive corporate data. This development addresses critical challenges in multi-agent AI ecosystems, facilitating cooperative optimization while respecting privacy and security constraints across organizational boundaries.[5]
Source: Fujitsu Official Release