Exaforce AI SOC Faces 20 Tough Questions on Agent Reliability, Permissions and Accountability
ScanNetSecurity editor Ueno Noboru conducted an in-depth interview with Exaforce's Shang Li, grilling the vendor on whether its Agentic SOC can truly be trusted with triage, investigation and response. Exaforce, founded in 2023 in San Jose by former F5 and Volterra engineers and backed by $200 million, built its platform from scratch around a three-layer architecture: raw event ingestion, a real-time knowledge graph, and LLM-driven agents. Li detailed how the system reduces alert handling time from hours to minutes, learns from analyst notes, and supports both fully autonomous and human-in-the-loop modes via Slack and Teams. The discussion covered prompt injection defenses, auditability of AI reasoning, data residency in Tokyo, and the reality that vendors cannot assume liability for missed incidents. Ueno also explored long-term impacts on junior analyst training and the company's roadmap toward greater automation of containment actions.
ScanNetSecurityNov 8, 2026Translated from Japanese