Who Can Use Your Data When Your Partners Need It?

Your partner needs your data. But do they need to see it?

That question changes what data collaboration can look like.

Scotiabank explored how AI could analyse money flows across banks to identify patterns associated with money laundering and flag potential instances of human trafficking. The participating institutions had a problem that many industries recognize: the data needed for the analysis was sensitive, and the organizations holding it had strong reasons to keep it protected.

The analysis could still happen without giving each participant direct access to the others' underlying data. Microsoft describes the approach as a multi-party data analytics scenario using Azure Confidential Computing and Opaque.

Why Does a Partnership Usually Mean Someone Has to Go First?

Much of the data sharing between companies still follows a familiar pattern: one organization sends another a copy of the information it needs, and a contract defines how that copy should be handled.

That approach can work for routine, low-risk information.

The calculation changes when the data itself is commercially sensitive or regulated - a bank's customer records, a retailer's purchase history, a manufacturer's supplier pricing, or a healthcare provider's patient records.

Consider a retailer that wants a bank to assess which customers may qualify for financing.

The bank needs information about purchasing behaviour to make the assessment. The retailer has a reason to protect that information. The bank has reasons to protect its own models and decision criteria.

The partnership now has two valuable assets to protect.

Historically, the choice has often been simple: share the data and accept the risk, or keep the data and accept a less useful analysis.

Confidential computing introduces another option.

What Changes When a Partner Can Use Data Without Seeing It?

The important change is where the computation takes place.

Instead of moving the underlying data to a shared environment where another organization can inspect it, the analysis can run inside a protected computing environment.

Microsoft's multi-party data analytics examples describe confidential computing as a way to protect data and models even while they are being processed. In the Scotiabank example, Azure Confidential Computing and Opaque were used to analyse interbank money flows for money-laundering detection and to flag potential human-trafficking cases.

This changes the permission model.

Instead of giving a partner broad access to a dataset, an organization can authorize a specific computation to use protected data for a defined purpose.

The underlying records remain protected inside the confidential computing environment, while the collaboration produces the result the participating organizations need.

The same principle can apply to the retailer-and-bank example.

The bank's model could evaluate protected purchasing information inside a controlled environment, with the collaboration designed to return only the agreed result rather than expose the underlying customer records.

The exact architecture depends on the use case, but the principle remains the same:

Give the computation access to the data it needs. Keep the underlying data protected.

What Does This Look Like Outside Banking?

The pattern extends beyond financial services.

Novartis Biome used a BeeKeeperAI partner solution running on Azure Confidential Computing to identify candidates for rare-disease clinical trials. Microsoft describes the scenario as a way to work with sensitive healthcare data while keeping the underlying datasets protected.

The same challenge appears in manufacturing.

A manufacturer and supplier may want to analyse shared production or capacity data. A group of logistics companies may want to identify broader routing patterns. Several organizations may want to train an AI model using data that none of them is prepared to place in a common repository.

The valuable insight comes from combining information.

The underlying data remains sensitive.

Confidential computing can provide a technical boundary in which the agreed analysis runs while the data remains protected during processing. Microsoft describes this as a way to enable multi-party analytics where competitive concerns, regulation, data location or privacy requirements make conventional data sharing difficult.

Who Can Use Your Data When Your Partners Need It?

The answer doesn't have to be “only us.”

It also doesn't have to be “anyone we send it to.”

A more controlled model is possible:

A specific computation gets permission to use specific data for a defined purpose, inside an environment designed to protect both the data and the computation.

That changes the starting point for a partnership.

Instead of asking:

“Which data can we safely give them?”

the better question becomes:

“What does the partner actually need to calculate, and can we let that calculation happen without exposing the underlying data?”

That distinction becomes increasingly important as companies collaborate across organizational boundaries while working with regulated, commercially sensitive and AI-relevant data.

The next data-sharing agreement may still require a file to move.

But before creating another copy of sensitive information, ask a different question:

Does your partner need your data - or do they need what your data can tell them?