Akash Network, a decentralized marketplace for renting spare computing power, on Tuesday said it had added "confidential compute," a way to run AI and other jobs so that the data stays hidden even from the operators whose machines are doing the work.

The system relies on so-called trusted execution environments, or TEEs, which are like sealed-off portions of a computer processor that insulate data from the host operating system, according to a blog post by Greg Osuri, co-founder and chief executive of Overclock Labs, the primary developer behind Akash.

The goal is to address the privacy concerns that arise when sending data jobs to a decentralized network of computers. The issue has proven a major blocker for institutional data users like medical centers, which otherwise might migrate over to take advantage of far lower prices.

"This shift fundamentally alters the security dynamic, finally making decentralized compute accessible for sensitive, enterprise-grade workloads that require absolute data privacy," Osuri wrote.

Akash's basic premise is that decentralized networks can be used to redistribute spare computing capacity that might otherwise go unused, say from a university with big servers that aren't used much on weekends.

Theoretically, the extra supply could help satisfy the seemingly insatiable computing demands of AI users. In April, DeAI News reported on Akash's Homenode program, which lets people plug home computers into the network to spread the work across countries and blunt the risk of a data-center outage or attack.

Inference capital markets

Such systems could play right into the "inference capital markets" that crypto financiers and Wall Street bankers are developing around AI computation, positioned to become a widely traded commodity, slung by hedgers and speculators like a barrel of oil or pork bellies.

In crypto markets, however, early excitement in the projects has given way to a more realistic assessment of the adoption timeline, especially with digital-asset markets in a slump; Akash's AKT token is down 67% this year alone, for a market value of $126 million, according to the pricing site CoinGecko.

Osuri noted that Akash is working to remedy a crucial gap that remains in the confidential setup: a final step called "attestation" where proof is provided that the TEEs themselves weren't compromised.

"Because attestation is mandatory for regulated applications, we want to be straightforward instead of overstating TEE capabilities," Osuri wrote. "We are currently developing tooling to integrate attestation as a smooth, integrated component of the deployment lifecycle."

Open-source alternative

Osuri has emerged as a leading advocate for open-source software and decentralized infrastructure, testifying on the topics before Congress in 2025. He co-founded Overclock Labs, the company that created Akash, in 2015, and the main network launched in 2020.

Akash bills itself as an open-source alternative to cloud-computing giants like Amazon, Microsoft and Google.

Keeping data private on a decentralized network is a problem other projects have approached, from many angles. In March, DeAI News reported on Nillion, whose nilGPT chatbot splits and encrypts data across many machines so no single one ever sees a full prompt, rather than relying on a trusted chip.

(HOW AI WAS USED IN THE PRODUCTION OF THIS PIECE: The first draft of this story was written on Claude Desktop using a custom skill for rewriting company blog posts, augmented with background research on Akash, Greg Osuri, the company's fundraising, and prior DeAI News coverage. I then used our Distro Publisher MCP connector to file the story as a draft to DistroVerse, our web app. I used the native (non-AI) story editor in DistroVerse to edit the piece, in this case a complete rewrite. I checked all the facts, during which I ended up adding a lot of attribution and clarifying several statements.)