IBL News | New York
Perplexity open-sourced the software it has been using to combat rogue AI agents, called SPACE (Sandboxed Platform for Agentic Code Execution).
Available on GitHub, SPACE is a sandbox platform for running code in isolated Firecracker microVMs.
Applications use its HTTP API to create Linux environments, execute commands, collect output, and manage sandbox lifecycles.
SPACE is useful when an AI application needs a separate execution environment for an agent, a user workspace, or a code-running job. The application decides what to run and which results to keep; SPACE manages the sandbox, its processes, and the connection that carries command output.
Perplexity said it lined up partners, including CoreWeave, Akamai and Crusoe, which agreed to offer SPACE to customers.
Sandboxes, or isolated environments for running AI models, have become a useful tool amid reports of agents escaping their constraints, finding their way onto the open web, and hacking prominent organizations like Hugging Face and the Australian government.
While AI companies focus on “alignment,” or training models to behave themselves, the security industry has been increasingly advocating for building strict guardrails around AI models.
Perplexity said it tested SPACE by giving nine frontier models instructions to try to break out. In 118 tries, none of them did.
Nvidia also released an open-source security tool, OpenShell, which runs AI agents in a sandbox and limits what they can reach.
Perplexity said SPACE also helps reduce wasted compute, which can be a byproduct of strict sandboxing.