Perplexity launches secure sandbox to make its AI agents secure and powerful

Perplexity AI Inc. today introduced a new feature that takes its current agentic artificial intelligence service, Computer, to perform better with greater security.

The company introduced SPACE, a sandbox platform designed to allow its AI agent to act with its full capabilities, while providing the highest level of security for agentic systems.

Perplexity Computer can be best described as a “digital worker” provided by Perplexity on top of its AI search engine. It takes user-described natural language goals and breaks them down into multi-step tasks, then orchestrates over 19 different frontier models – including Claude, Gemini and Grok – to interact with tools including web search, email, Notion, Slack, data curation and more.

In this manner, it is similar to agentic AI work systems released by other companies, such as Anthropic PBC’s Claude Cowork and OpenAI Group PBC’s ChatGPT Work.

The system can recall past work, the person’s specific style and allows the user to walk away and let it run on its own. It can also be set up to run on a schedule, and monitor and activate when data changes over hours, days and months.

Work that lasts over hours, days and months sometimes requires restarts. To handle this, SPACE wraps sessions that can be paused, resumed or branched into multiple sandboxes. Fitting the name, a sandbox is an isolated space where all the data an agent is working on can live, serving as a secure context safely packaged away from other agents or systems.

According to Perplexity, the company began by evaluating numerous off-the-shelf sandbox systems. Many setups may be good at isolating, but leave user passwords, keys and other information exposed. Others are good for short-lived jobs, but lead to confusion for sessions that must run for hours or days.

The company designed SPACE to solve both the security problem and the long-term capability while also providing something that could be built and torn down within moments.

Building a secure environment for agentic AI

Every task runs within an instance of Firecracker microVM, an open-source virtual machine developed by Amazon Web Services Inc. and designed specifically to improve resource utilization for AWS services such as Lambda and Fargate. Traditional virtual machines provided isolation but are too slow to boot and consume too much memory for Amazon’s purposes; on the other hand, containers are extremely fast but lack hardware-level isolation. Firecracker created a happy medium by booting tiny virtual machines in as little as 5 milliseconds with minimal overhead.

Perplexity added that SPACE uses this technology to allow it to provide a full set of controls: combining pause, resume, session forking, per-session credential isolation, forwarding and orchestration into one platform.

The end result is that consumers have a system that captures live memory along with files that can be updated as frequently as every minute. It can archive and retain information for up to a week. This means that a user can walk away mid-task, come back a week later, and the agent will be able to start right where they left off.

As for passwords, keys and other sensitive information, none of that is stored within the sandbox and remains entirely in the control of the user through their own password and key management systems, and is only accessed when needed. Users also bring their own encryption keys for the information stored in the sandbox. If a user revokes the encryption key, new sandboxes cannot read the data, making it completely secure.

SPACE capabilities are live in Computer today. The company added that it’s designed to run anywhere. In early testing using Nvidia Corp.’s Vera architecture, Computer-style workflows ran around 1.5 times faster than current production references and launched concurrent sandboxes 1.9 times faster.

The company added that it has been using SPACE internally for the past two months. Over the past week, it supported more than 1.25 million sandbox creations and 11.9 million sandbox reconnects.

Perplexity said SPACE is only the first iteration of the secure environment that it intends to develop. As AI and agentic workflows continue to evolve, the company said it will update its infrastructure to keep pace with innovation.

Image: SiliconANGLE/Microsoft Designer

A message from John Furrier, co-founder of SiliconANGLE:

Support our mission to keep content open and free by engaging with theCUBE community. Join theCUBE’s Alumni Trust Network, where technology leaders connect, share intelligence and create opportunities.

15M+ viewers of theCUBE videos, powering conversations across AI, cloud, cybersecurity and more

11.4k+ theCUBE alumni — Connect with more than 11,400 tech and business leaders shaping the future through a unique trusted-based network.

About SiliconANGLE Media