Confidential Computing

Confidential Computing is a technology that keeps data hidden in encrypted form even at the exact moment a computer is performing calculations on it. Normally, a computer has to decrypt data and load it into memory to compute on it, and that moment is precisely when the data is most exposed to leaks. Confidential Computing creates a sealed room at the hardware level, so that even the cloud company that owns and rents out the computer cannot see what calculations are happening inside that room.

Here's an analogy: imagine two companies that don't trust each other, each bringing documents locked in their own safe into a sealed room that no one else can enter. They compare the documents inside, and only the final result leaves the room. Neither company, nor the third party managing the room, can see what documents were exchanged inside — only the final output is passed outside.

This is why it's used in situations where parties need to collaborate while each keeping their own secrets protected. A representative example is AI model evaluation, where one side wants to keep test questions secret and the other wants to keep the model's internal structure secret.

The article states, "For this pilot, Google used 'Confidential Space,' part of Google Cloud's confidential computing portfolio." A common point of confusion: Confidential Computing isn't the name of a specific product but a technology category covering many services, and Confidential Space is a specific Google Cloud service within that category.