Patricia Horotan, Founder & Chief Strategy Officer, BlueMatrix
Who owns the question?
For as long as investment research has existed, its producers have wanted to know what happens to it after publication.
Who reads it? Which analysts carry influence? Does a report shape an investment decision, or disappear into the vast flow of information passing through an investment firm every day?
AI has made these questions more urgent. Research is increasingly read by machines, and the banks producing it are understandably reluctant to see their work disappear into systems they can neither observe nor control.
A research report represents considerable intellectual effort. It contains an analyst's judgment, a department's accumulated knowledge, often years spent following a company or an industry. Once that report enters another firm's AI environment, the producer may have little visibility into what happens next. Its arguments can be extracted, its conclusions incorporated into answers, its author's identity forgotten.
The sell side has good reason to worry.
But there is another side to the question of intellectual property in investment research, one that has received far less attention.
In conversations with investment firms about research access, a familiar tension emerges between the sell side's desire for greater visibility into research consumption and the buy side's reluctance to disclose information about its investment process. There is growing interest in the possibility of a reciprocal exchange of information, but these discussions also reveal a distinction worth examining.
Investment firms increasingly regard their readership, their selection of research, and the ways they work with that research as intellectual property in their own right.
An investment bank owns the research it produces. But an investor's decision to read one report rather than another, to question a particular assumption, to compare two analysts' conclusions, or to revisit a forecast made three years earlier is part of the investor's own intellectual activity.
The report belongs to its author. The inquiry belongs to its reader.
Historically, the distinction was largely preserved by the nature of research distribution. The sell side could measure readership and infer influence, but the investor's reasoning remained private.
AI is changing that separation.
One proposed solution to the sell side's concerns about losing control of its content is to make research available through MCP, the Model Context Protocol, rather than allowing it to be ingested wholesale into the buy side's models.
The attraction is understandable. Instead of surrendering an entire research library to another institution's AI environment, a publisher can make its content available through controlled tools. Access can be governed, entitlements enforced, and usage observed.
But the server can potentially observe the investor, too.
When an investor's AI system queries an externally operated server, the operator can see the requests it receives. Depending on how the service is designed, those requests may reveal far more than which reports were accessed: which companies are attracting attention, which assumptions are being challenged, which historical predictions are being revisited, and which questions the investor keeps asking.
The server need not see the investor's original prompt or the reasoning taking place inside its models. The requests crossing the boundary may reveal enough.
The technology intended to restore the research producer's visibility may also expose the research consumer's thinking.
The sell side worries about losing control of its intellectual property. The buy side has reason to worry about losing privacy over its intellectual inquiry.
These are not competing claims to the same property. They are different interests, arising on opposite sides of the same relationship.
Nor is this exclusively an MCP problem. Conventional APIs can expose similar information. What matters is where the inquiry takes place, what crosses institutional boundaries, and who is permitted to observe and retain it.
As AI makes research consumption more revealing, the distinction between measuring influence and observing inquiry becomes increasingly important. Producers have a legitimate interest in understanding how their work is used. But knowing that research has influenced an investment decision is quite different from knowing what the investor is trying to discover.
The challenge is to protect both.
The buy side may be more receptive to an exchange of information than has traditionally been assumed. Its reluctance to share research consumption data is not absolute. It depends on what is being disclosed, what is received in return, and who can be trusted to manage the exchange.
And the buy side has interests of its own. Information about changes in analyst coverage or patterns of research activity may be valuable alongside the published research itself. But increasingly, investors also want to know something more fundamental about the research they consume: whose judgment does it represent?
As AI-generated content proliferates, that question becomes harder to answer. Investment firms building proprietary models around the knowledge and experience of the world's best analysts have little interest in filling those models with synthetic material of uncertain origin.
The ability to establish provenance and distinguish authenticated human contribution from machine-generated content is therefore becoming part of the value of research itself.
The research producer wants its intellectual property protected and its contribution recognized. The research consumer wants its inquiry kept private and the material entering its systems to be trustworthy. Both may be willing to exchange information of value, provided neither has to surrender what matters most.
For years, research distribution has been concerned with getting the right content to the right recipient. In an AI environment, distribution alone is no longer enough. The infrastructure must establish the provenance of what is delivered, preserve the rights attached to it, measure its use without unnecessarily exposing the user's intentions, and govern the information exchanged between the two sides.
Neither party should have to depend entirely on the other's discretion.
At BlueMatrix, we have spent years building the infrastructure through which investment research is authored, governed, and distributed. Much of our recent work has concerned the preservation of attribution, entitlements, and provenance as research enters AI systems.
That responsibility now extends further.
Protecting the intellectual property of the producer is essential. So is protecting the intellectual privacy of the consumer. The two should not be treated as a trade-off.
The research industry has always depended on a relationship between those who develop ideas and those who put them to work. AI is making that relationship more complicated, but it is also giving us an opportunity to define it more carefully.
We have spent a long time asking who owns the research, and more recently, who owns the answer.
Perhaps it is time to ask who owns the question.