Security · November 14, 2026
Also published in Português (Brasil)
Akamai Introduces Agent Sandbox Isolation System to Prevent Data Leakage from AI Agents
Akamai has launched a new isolation system called Agent Sandbox to address data leakage risks associated with autonomous AI agents. This initiative was announced during a media briefing held via Zoom on November 12, 2026. The development comes as companies face increasing operational costs due to high token consumption in artificial intelligence (AI) applications.
During the briefing, Chief Technology Officer of Akamai Cloud Computing, Jay Jenkins, explained that many companies using commercial AI models for software development often incur rising token costs as their operational scale grows. To counter these expenses, Akamai's solution includes an open endpoint compatible with industry standards, allowing businesses to shift part of their workload to open-weight models to reduce expenditures.
Jenkins emphasized that organizations can build applications, architectures, and coding while utilizing open-weight models for supporting tasks such as unit testing, documentation, and code review. This approach is expected to significantly reduce generative AI usage costs.
In addition to cost efficiency, security within the Agent AI ecosystem has emerged as a primary concern. With autonomous AI agents capable of independently writing code without constant human intervention, cybersecurity risks have increased. The Akamai Agent Sandbox is designed to run AI agents in an isolated environment using WebAssembly technology developed in collaboration with the Cloud Native Computing Foundation.
This innovative system aims to provide flexibility through serverless inference services, seamless API compatibility, and stringent protection via the Agent Sandbox, enabling industries to innovate without facing security risks or heavy reliance on a single service provider.