Temporal Acquires Security Expertise to Protect AI Agents in Enterprise Systems
Temporal, a Bellevue-based infrastructure company valued at $12.55 billion, is integrating security controls for AI agents by bringing on the team from Oso, a New York startup specializing in application authorization. The move addresses critical vulnerabilities where autonomous agents must access databases and external systems while maintaining proper security boundaries and preventing unauthorized actions. By combining Oso's access control frameworks with Temporal's durable execution technology, the company aims to ensure agents operate within approved permissions and cannot bypass security guardrails.
Temporal's acquisition of Oso's team represents a significant expansion into security infrastructure for autonomous systems. The Seattle-based company, which has roughly doubled its workforce to 570 employees over the past year, is leveraging its existing durable execution platform—technology that preserves application state across failures—to embed authorization controls directly into agent operations. This integration ensures that when AI systems interact with databases and external tools, they remain confined to predetermined permission boundaries and cannot repeat previously rejected actions.
The deal follows Temporal's explosive growth trajectory, with the company reaching $250 million in annualized revenue and expanding its enterprise customer base to include major firms like JPMorgan Chase, Netflix, and NVIDIA. By combining execution reliability with access control frameworks, Temporal positions itself at the intersection of AI capability and enterprise governance—a critical layer as companies move toward deploying autonomous agents in production environments.
This development may shape how enterprises deploy AI agents across sensitive systems. By embedding security controls into infrastructure rather than relying solely on model-level safeguards, the approach could reduce risks associated with unauthorized data access or system manipulation. However, the effectiveness ultimately depends on how comprehensively these controls address evolving agent capabilities and attack vectors. Organizations implementing such systems face questions about permission configuration complexity and whether controls keep pace with AI advancement, potentially affecting adoption timelines and security outcomes across industries.