Whose are the prompts you use at work?

The attached document from the Spanish Data Protection Agency analyzes in DEMÓCRATA the legal and ethical limits of companies’ access to their employees’ prompts in AI tools

3 minutes

Add DEMÓCRATA to Google

Published

3 minutes

Corporate control has a fairly intuitive limit: in general, the company can reasonably control the way its employees use the work tools that it places at their disposal, since they are considered to be the property of the organization.

The appearance of the artificial intelligence begins to blur the boundary of that legally recognized power granted to the employer and leads us to ask ourselves whether the company can access, without restrictions, the prompts that its employees enter into a corporate artificial intelligence.

The answer could vary depending on whether we consider that a prompt is a simple instruction given by the employee to the AI, or if it is something more, in the sense that it can contain a business strategy, a legal hypothesis, a technical solution, confidential information, or personal data. But it can also reveal something much more difficult to protect: what is the reasoning that develops that person when doing their work.

The answer could vary depending on whether we consider that a prompt is a simple instruction given by the employee to the AI, or if it is something more

Today, this debate can no longer be described as futuristic, and it seems destined to be a source of labor controversy in the very near future, if it isn’t already.

To illustrate this argument, let us remember that, in 2026, Meta had to temporarily suspend a program intended to collect information about the activity of its employees—including mouse movements, keystrokes, and other interactions—in order to train artificial intelligence systems.

This case reveals an evolutionary trend in technologies, since certain uses that allow us to control what a worker does can -as well- become a way to record how they work and -potentially- how they think.

That use of AI in the workplace introduces another reasonable doubt about the ownership of the skills of working people: what happens when that professional knowledge is recorded in thousands of prompts?

Let us imagine a company that requires its employees to use exclusively a corporate AI tool. The tool keeps the histories. The systems department can access them. Human Resources finds out that it can obtain information about productivity, knowledge, work methods, and even about the difficulties each employee encounters or any potential work or psychological problems of those employees.

Could the company use such information? If so, for what purpose?

Although the Workers’ Statute recognizes employers’ powers of control over the employee, it also sets limits linked to privacy and data protection. That is to say, ownership of the device does not equate to an unlimited right over everything that is done with it.

Additionally, the difficulty of this issue is also perceived in another type of regulation, as is the case with that relating to business secrets, in which, in its article 1.3, a distinction is reached between the protected knowledge of the company and the experience and competencies acquired honestly by the worker during their professional career.

Indeed, a worker may have developed over years a specific methodology for analyzing contracts, designing products, or solving technical problems. If that methodology is progressively reflected in their interactions with a corporate AI, the company (but also the AI) can have available an extraordinarily valuable record of that knowledge.

Who does it belong to? What are the limits on its use?

There are no shortage of signs that the conflict will arise. The American Bar Association has already opened the debate about whether the prompts and AI responses can come to have legal consideration as “records” that certain organizations in certain sectors are required to keep.

Although this may be an aspect closer to those known as neurorights, in the business sphere too terms such as “cognitive privacy” are beginning to be coined in light of the possibility that new technologies not only allow information about an employee to be known, but also allow aspects related to their thought process to be deduced.

At this point, we must differentiate two assumptions. One, consisting in the fact that the employer wants to know whether a worker uses an authorized AI application (which may be considered a security measure because with it it diligently protects its secrets, avoids leaks, and guarantees safe use of these tools). On the other hand, where the company wants to access systematically everything that its employees ask the AI, which could be considered a type of cognitive surveillance whose legality may raise doubts.

Additionally, we can identify another side effect: if workers know that every question they ask can be read by their superiors, they will probably start behaving differently. Or perhaps they will ask less, so as not to seem that all the work is done by the AI.

Thus, the tool that was supposed to boost their creativity can end up conditioning it.

Until now, we were concerned that artificial intelligence might replace our work. From what has been seen, perhaps we can conclude that, if it does, it won’t be so much because it knows what we do, but because it will have learned how we do it first.