Akamai Agent Sandbox Gives Enterprise AI Agents a Safe Place to Run

Akamai has launched two new products for its AI stack on Akamai Cloud, Akamai Inference and Akamai Agent Sandbox, in a move aimed squarely at enterprises putting AI agents into production. According to Akamai, the October 27, 2026 announcement gives companies a single platform where agents can be served, run, and secured close to the users and data they depend on.

The launch matters because the hardest part of deploying agents is no longer the model. Enterprises moving AI into production need serving infrastructure, access to their own data, routing and governance, and, as agents begin to write and execute their own code, a hard boundary around what that code can reach. Akamai Agent Sandbox is designed to supply that boundary, while Akamai Inference supplies the managed model serving underneath it.

What Akamai Agent Sandbox does

Akamai Agent Sandbox gives developers a secure place to run agents and the code those agents write. According to Akamai, agents run on Akamai Functions, while the code they write runs in its own isolated environment, sequestered from the agent runtime. Delivered as an open source component of the Spin framework, the sandbox uses WebAssembly isolation to set clear limits on compute, memory, outbound connections, and system capabilities.

That separation is the point. An agent that books travel, updates records, or writes a script on a user's behalf should not be able to reach beyond its brief simply because it generated code. Developers can deploy agents with confidence about where they run and what they can reach, according to Akamai, because the sandbox enforces those limits at the execution layer rather than relying on the agent to behave.

Akamai positions the sandbox within a broader choice of isolation models. Agent isolation is converging on several approaches, from microVMs with snapshot and fork capability to policy-enforced containers to WebAssembly-isolated functions, and Akamai lets developers choose the architecture that fits their workload. Options include NVIDIA-accelerated and dedicated CPU infrastructure, managed Kubernetes, and Akamai Functions for serverless execution across edge and core locations. The company also plans to add capabilities that govern what agents can reach and what can reach them.

Akamai Inference completes the stack

Akamai Inference gives enterprises managed access to open-weight and custom models through OpenAI-compatible APIs, so existing applications can move without rewrites. Customers can choose serverless, pay-per-use inference or dedicated deployments. At launch, the catalog includes 12 open-weight models ranging from 8 billion to 2.8 trillion parameters, including NVIDIA Nemotron 3 Ultra, OpenAI gpt-oss-120b, and Google Gemma 4, along with embedding models for retrieval-augmented generation.

Akamai Cloud provides managed databases and object storage and connects to private and public cloud platforms, so models can draw on enterprise data. Akamai Inference then handles orchestration, model execution, and authentication. Requests run on the Akamai global network, which reduces latency and helps ensure high availability and consistent performance while maintaining predictable service levels, according to the company.

The infrastructure underneath spans dense GPU and CPU clusters in core regions and more than 4,350 edge locations, protected by the same security portfolio that defends many of the world's largest online businesses. Akamai says that foundation has already secured multiyear contracts for Akamai Cloud collectively worth more than 14 billion dollars, a sign of how quickly managed AI infrastructure has become a board-level purchase.

Why governed agent execution is now the bottleneck

The Akamai launch lands in a week dominated by agent governance. IBM unveiled governed identity and credential controls for enterprise agents at TechXchange 2026, as reported by GenZ NewZ, reflecting the same concern from a different angle: who owns each agent, what it can access, and how its actions can be traced. Akamai Agent Sandbox answers the execution half of that question, while identity and access controls answer the permission half.

Testing is the other half of trust. Independent assurance for agents is also emerging, as covered in GenZ NewZ coverage of agent assurance testing. Together, identity controls, assurance testing, and isolated execution form a practical trust stack for agents that do real work inside companies, rather than merely drafting text.

Akamai executives framed the launch in those terms. Adam Karon, chief operating officer and general manager of Akamai Cloud, said customers proved the thesis that AI inference would represent the majority of enterprise focus and investment, and pulled the company further. According to Karon, customers want to build, run, and protect AI on the same platform, and the new products add the inference and execution layers that turn Akamai infrastructure into a full AI stack.

GPU muscle behind the launch

Akamai also expanded its dedicated GPU fleet to include NVIDIA Blackwell Ultra GPU clusters and is preparing its infrastructure to support the NVIDIA Vera Rubin platform in 2027. The fleet now spans NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs for low-latency inference and Blackwell Ultra clusters for enterprises that want to build, tune, and self-host models on infrastructure they control.

Each eight-GPU Blackwell Ultra node carries 288 GB of HBM3e memory per GPU, 2.3 TB in total, enough to hold frontier-class models. According to Akamai, the architecture targets the workloads driving that memory demand: test-time scaling, chain-of-thought reasoning, and long-context agentic work. GPU clusters give enterprises control over model versions and deprecation timelines and run without rate limits or shared-tenancy queueing.

Jon Alexander, senior vice president of cloud technology at Akamai, said enterprises should not have to choose between speed and control when they put AI into production. Some teams want to ship an AI feature this quarter on a managed platform, while others need dedicated GPUs to run frontier-class open models on their own terms, and every team running agents needs a boundary around the code those agents write.

Availability and what to watch

Akamai Agent Sandbox, via the Spin SDK, and Akamai Inference are available in beta, according to Akamai. The Inference beta includes serverless deployment options, and customers are directed to contact Akamai Sales for NVIDIA B300 GPU cluster availability.

The questions worth watching are practical: how quickly developers adopt the Spin-based sandbox versus container and microVM alternatives, whether the OpenAI-compatible API genuinely removes migration friction, and how the promised governance controls for what agents can reach take shape. For enterprises, the launch is a reminder that agent strategy now has three layers, not one: the model, the identity, and the sandbox. Akamai is betting that owning all three on one cloud is what turns agentic experiments into production systems.

Sources: Akamai announcement via GlobeNewswire; Akamai Cloud AI infrastructure details from the same announcement.