What if companies could finally use their most valuable data?
Sifted
Companies have data too sensitive for outside AI, so it stays locked up. Now tools like confidential computing could let them use it without handing it over.
Based on reporting by Sifted — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Every company has a stash of information it treats like a secret weapon. Banks have transaction histories. Drugmakers have years of research. Cybersecurity firms have threat data that shows where the bodies are buried, so to speak. And that is exactly the data many businesses have been afraid to feed into outside AI systems.
The fear is not abstract. Put sensitive customer records, intellectual property or regulated data into an AI service and you immediately have questions about access, processing and what happens after the model has done its work. For companies dealing with GDPR, industry rules or national security demands, those questions can carry real consequences. So while businesses pour money into AI, some of their most valuable data sits on the sidelines.
That tension is where confidential computing comes in. Instead of protecting data only when it is stored or moving between systems, it protects it while it is being processed. The idea is simple enough: create a locked room for computation, where the data and the model can work together, but the infrastructure operator — and sometimes even the model provider — still cannot see what is inside.
VAST Data is betting that this will matter. Its new DataEnclave technology is meant to let companies bring AI models into confidential environments that already contain sensitive data, while keeping the model, the data and the underlying infrastructure separated from one another. The company’s pitch is a reversal of the usual setup. Don’t send the data out to the model. Bring the model to the data.
That could change the economics of enterprise AI. A bank could analyze customer information without handing over the underlying records. A cybersecurity company could run AI over highly sensitive threat data. Governments could use AI on information they would never put into a public service. And for European startups built on specialized datasets in finance, healthcare, defence and industrial technology, it could reduce the old trade-off between using powerful AI and protecting the thing that makes the business valuable.
None of this makes governance optional. Companies still have to decide what data to use, which models to trust and what employees or agents can do with the output. But it does suggest the next real AI gain may not come from a flashier model. It may come from finally using the data companies have been guarding all along.
My take — AI-written commentary, not fact-checked reporting
This is the part of AI that actually matters: not another victory lap for bigger models, but a way to use the valuable stuff companies already own. The hype machine loves shiny benchmarks; businesses care about not handing their crown jewels to a random API. Confidential computing looks less like magic and more like common sense, which is usually how the useful tech sneaks in.
Read more about this at: Sifted