When the App Becomes the Employer
The €825 million Uber fine reveals how platforms can exercise institutional power without appearing to employ anyone
The Black Box Divide
SIGNAL | Uber presented itself as a platform connecting drivers and passengers. But when its systems could evaluate behaviour, identify suspected fraud and block access to work, the app began exercising employer-like power. Europe is now asking whether such power may exist without a recognisable human decision-maker.
The €825 million fine imposed on Uber by the Dutch Data Protection Authority concerns GDPR violations surrounding automated deactivation. Uber disputes the findings and has appealed.
Yet behind the legal dispute lies a deeper institutional shift: an interface can replicate the power of an employer without necessarily assuming the responsibilities of one.
When deactivating an account amounts to economic dismissal, system administration becomes governance.
🟦 When does a platform stop being an intermediary?
Uber has long presented itself as a technology platform connecting independent drivers with passengers.
Yet an intermediary normally facilitates transactions between others. It does not necessarily monitor performance, interpret customer ratings, identify suspected misconduct and determine whether one party may continue participating in the market.
The case originated with complaints from more than 170 French drivers. According to the regulators, Uber’s systems could temporarily disconnect drivers suspected of fraud. Drivers affected by persistently low ratings could also lose access to the platform.
Once a platform performs those functions, it begins to resemble more than a marketplace. It becomes the workplace, the monitoring system and the gatekeeper at the same time.
That does not automatically determine whether a driver is legally an employee. It does reveal that the platform can exercise employer-like power without necessarily accepting the institutional identity and responsibilities of an employer.
The platform monetises the flexibility of an open market while disciplining participants with the unilateral authority of a manager.
The market remains open — until the platform decides that someone may no longer enter it.
🟦 Is losing access to an app different from losing a job?
Technically, Uber did not need to dismiss anyone. It could change an account status. Operationally, the distinction may mean very little to the driver.
When access to the platform is also access to customers, work and income, deactivation is not merely a digital inconvenience. A technical restriction produces an immediate economic consequence.
This is one of the defining characteristics of platform power. Decisions with profound effects can be presented as routine system administration.
A driver is not necessarily told that an institution has judged them. The screen simply changes. The language becomes softer as the power becomes harder.
An account is “restricted”. Access is “temporarily unavailable”. A profile is “under review”. Behind those neutral formulations, someone may suddenly be unable to earn.
The platform therefore does more than organise a market. It determines who remains visible, who may continue participating and under which conditions. What appears in the interface as deactivation may be experienced in the economy as dismissal.
🟦 Does adding a human make an automated decision human?
Article 22 of the GDPR gives individuals the right not to be subjected to decisions based solely on automated processing when those decisions produce legal or similarly significant effects.
The protection is not absolute. European law permits exceptions, but it also requires safeguards, including opportunities for human intervention and contestation. Those protections are set out in Article 22 of the GDPR. But adding a person to the chain does not guarantee human judgement.
If a reviewer merely confirms an algorithmic flag without sufficient time, context or authority to overrule it, the human is not functioning as a safeguard. The person has become part of the automation.
Meaningful intervention requires someone who can examine the evidence, identify errors, consider circumstances the system could not see and reverse the outcome.
Otherwise, “human in the loop” becomes procedural theatre. The system recommends. The employee confirms. The organisation calls it oversight.
🟦 Is this still merely a privacy case?
The fine was imposed under the GDPR. It is therefore tempting to classify the dispute as another European data-protection case. But privacy is only the legal doorway through which a much larger issue becomes visible.
Data is no longer collected merely to describe people. It is increasingly used to classify, rank, predict, permit and exclude them.
A fraud model can interrupt a driver’s income. A credit model can prevent someone from obtaining a mortgage. A recruitment system can ensure that a candidate is never seen. An insurance model can determine whose claim is treated as suspicious. A public-sector algorithm can influence access to benefits, housing or care.
The essential question is therefore not only whether personal data was processed lawfully. It is what kind of authority an organisation acquires once that data becomes the basis for decisions about participation in society. The GDPR begins as a privacy law.
In cases such as Uber, it becomes a boundary on institutional power. The decisive shift begins whenever software moves from supporting a decision to exercising power over a person.
🟦 Is the algorithm really the black box?
Europe is often criticised for regulating technologies it did not build. The criticism is not entirely misplaced. Rules can become fragmented, burdensome and detached from the way organisations actually operate.
But the Uber case is not fundamentally about whether Europe approves of algorithms. Nor is it simply another confrontation between a European regulator and an American technology company. It concerns the architecture surrounding the system.
Who defines the rules? Who verifies the data? Who receives the warning? Who can reconsider the outcome? Who remains responsible when the system is wrong?
Technical explainability alone cannot answer those questions. Even a perfectly explainable model can sit inside an organisation that offers no meaningful route to challenge its power.
The black box is therefore not always the algorithm. Sometimes it is the institution that has placed the algorithm between itself and the people affected by its decisions.
The opacity is not necessarily an engineering accident. It can also function as an institutional buffer — separating the organisation exercising power from responsibility for the consequences.
Signal
The future of work may contain fewer offices, fewer conventional managers and fewer recognisable employment relationships. It will not contain fewer decisions about people.
Those decisions will increasingly be embedded inside platforms that observe behaviour, organise access and impose consequences while continuing to describe themselves as neutral intermediaries. The Uber fine makes that shift visible.
The case is not simply about an algorithm that may have made the wrong decision. It is about a system capable of exercising economic power without clearly presenting itself as the institution responsible for doing so.
Europe does not need to place a symbolic human beside every automated process. It does need to ensure that whenever access to work, finance, insurance or public services is withdrawn, someone remains capable of explaining, reconsidering and taking responsibility for that decision.
Because when the app becomes the employer, the bank officer, the insurer or the public administrator, “the system decided” is no longer a technical explanation. It is an admission that power has become difficult to locate.
This Signal is part of The Black Box Divide, Altair Media’s series on power, transparency and control in European artificial intelligence. The series examines what happens when intelligent systems become more capable while the decisions, institutions and responsibilities surrounding them become harder to see.
Credit
AI-generated editorial image by Altair Media
Caption
A technical change in account status can have the economic effect of dismissal. The Uber case raises a wider question: who remains responsible when access to work is controlled through an automated interface?
