SoftBank Develops AI Agent Security Tech for 2026 Launch

Tokyo, Japan - SoftBank Corp said it has developed a prototype technology that enables artificial intelligence agents from different organizations to communicate while maintaining separate data governance frameworks.

The telecommunications company stated that its "Agent Firewall" technology addresses challenges that arise when AI systems attempt to exchange data across organizations operating under different data handling rules and management standards. 

SoftBank has filed patent applications for the technology.

The company plans to commercialize the system around fiscal year 2026, following internal testing and collaboration with partners.

SoftBank stated that the development addresses anticipated obstacles in multi-AI agent environments, where enterprises have varying data location requirements, access control standards, and compliance rules that could create barriers to inter-organizational data collaboration.

The Agent Firewall incorporates four technical features, according to the company. Dynamic governance control automatically adjusts inspection intensity based on communication risk levels and required security standards.

The system provides authentication and authorization functions that permit communication only between trusted AI agents.

Contract and policy-based controls automatically detect violations in accordance with contractual conditions and organizational policies, thereby preventing data loss incidents.

The technology includes monitoring, auditing, and log management capabilities that record communications in tamper-resistant ledgers with real-time monitoring.

SoftBank stated that the technology complements existing protocols, including the Agent-to-Agent Protocol and Agent Communication Protocol, while enabling the consistent application of security policies regardless of the protocol type.

The system operates on zero-trust principles rather than traditional network perimeter security approaches, checking communications only when necessary and within the required scope to maintain security while reducing system load, the company said.

This design enables deployment across large-scale distributed network environments.

SoftBank stated that it will accelerate the development and verification processes to meet its 2026 commercialization target.

The company indicated that the technology is expected to integrate with sovereign cloud and sovereign AI systems for use by corporations, municipalities, and research institutions handling sensitive data.

SoftBank described the development as part of its ongoing research into advanced technologies designed to support secure and reliable AI infrastructure.