Algorithmic management: what the enforced record shows for gig workers in 2026
2026-09-02
Three European regulators fined delivery and ride-hail platforms for managing workers by software. Read together, the fines punish one thing: the missing human.
Built on verified case files. The argument below leans on evidence The Internet Ninja validated against the public record and published in full, method included.
For a gig worker, the boss is the app. It hands out the next order, scores the last one, and can switch off the account that pays the rent. When that decision goes wrong, there is often no person to appeal to, because there was no person in the decision.
That is not a complaint about the future of work. It is what three European regulators found and fined between 2021 and 2026. Italy’s Garante fined Deliveroo 2.5 million euros and Foodinho 2.6 million euros, and the Dutch authority fined Uber 824,990,000 euros (source), the second-largest GDPR penalty on record.
What algorithmic management means
Algorithmic management is the use of software, rather than human supervisors, to assign, monitor, score and discipline workers. The system decides who gets work, how they are rated, and when their access to earning is cut.
The label is not marketing. Italy’s data-protection authority used the concept directly, and TechCrunch reported that “algorithmic management of gig workers has landed Glovo-owned on-demand delivery firm Foodinho in trouble in Italy” (source).
What algorithmic management looks like in the workplace
The three cases below are the same mechanism at different scales. A platform replaces the supervisor with a model, and the model decides who works.
Deliveroo Italy ran order-assignment and shift-booking through software that tracked about 8,000 riders, capturing geolocation every 12 seconds and storing routes for six months. Foodinho scored and booked its riders the same way, across more than 35,000 workers by 2024. Uber went further: its software permanently deactivated driver accounts for persistently low ratings, cutting off income with, the Dutch regulator found, “no human assessment” (source).
The proof: what TIN verified
TIN audited each of these against the regulators’ own published orders. These are not vendor claims; they are adjudicated public records.
In Italy’s Garante fine of Foodinho, the authority fined the Glovo-owned platform 2.6 million euros in 2021 for failing to tell riders how the algorithm decided their work, then a further 5 million euros in 2024 over the data of more than 35,000 riders, and it banned the company’s use of riders’ facial-recognition data (source).
In the Garante’s parallel fine of Deliveroo, the regulator fined the platform 2.5 million euros for non-transparent order-assignment algorithms and ordered it to add human-review safeguards, granting 60 days to correct the violations and a further 90 to fix the algorithms (source).
In the Dutch DPA’s fine of Uber, the authority ruled that Uber “violated the prohibition of fully automated decision-making under the GDPR” and imposed 824,990,000 euros; Uber has stopped the practice and appealed (source).
The record in one table
| Platform | Regulator | Year | Fine | The core finding |
|---|---|---|---|---|
| Deliveroo Italy | Garante (IT) | 2021 | 2.5M euros | Non-transparent rider-management algorithms; ordered to add human-review safeguards |
| Foodinho (Glovo) | Garante (IT) | 2021, 2024 | 2.6M + 5M euros | Riders not told how the algorithm decided; later a biometric-data ban |
| Uber | Autoriteit Persoonsgegevens (NL) | 2026 | 824.99M euros | Automated account deactivation with no human assessment |
What does the record show about algorithmic management in the gig economy?
The easy reading is that these are privacy fines, and algorithmic management collided with the GDPR. That reading is true and it misses the load-bearing part.
Line the orders up and the punished defect is the same one every time, and it is not the algorithm. It is the missing human. Deliveroo and Foodinho were not told to switch the software off; they were ordered to explain it and to add a person who could review a contested decision. Uber’s record 824.99-million-euro fine was for “fully automated decision-making” specifically, the version with no human in the loop.
So the fines do not price the model. They price the absence of a route to a person who can be asked why, and of a worker who can contest the answer. An algorithm that assigns orders is cheap to run and, so far, cheap in law. An algorithm that ends someone’s income with no appeal is what regulators have started charging nine-figure sums for.
The bottom line
If you manage workers with software, the contestability path is the product, not the compliance afterthought. Build the route to a human who can review and reverse a decision before you ship the model that makes it, because the enforced record now says that is the part a regulator reads first. The lesson generalises past gig work: any automated decision that can take something essential away from a person is safe only to the degree a human can still be held to account for it.
Sources
- Garante per la protezione dei dati personali, “Rider: il Garante privacy sanziona Foodinho per 2,6 milioni di euro,” 2021-07-05, https://www.garanteprivacy.it/home/docweb/-/docweb-display/docweb/9677377
- Garante per la protezione dei dati personali, “Rider: sanzione di 5 milioni di euro a Foodinho,” 2024-11-22, https://www.garanteprivacy.it/home/docweb/-/docweb-display/docweb/10074840
- Garante per la protezione dei dati personali, “Rider: il Garante privacy sanziona Deliveroo Italy per 2,5 milioni di euro,” 2021-08-02, https://www.garanteprivacy.it/web/guest/home/docweb/-/docweb-display/docweb/9687860
- Autoriteit Persoonsgegevens, “Uber fined nearly 825 million euros for automated driver blocking,” 2026-08-21, https://www.autoriteitpersoonsgegevens.nl/en/current/uber-fined-nearly-825-million-euros-for-automated-driver-blocking
- TechCrunch (Reuters), “Uber faces fine of nearly $1B over automated driver suspensions,” 2026-08-23, https://techcrunch.com/2026/08/23/uber-faces-fine-of-nearly-1b-over-automated-driver-suspensions/
- TechCrunch, “Italy’s DPA fines Glovo-owned Foodinho $3M, orders changes to algorithmic management of riders,” 2021-07-06, https://techcrunch.com/2021/07/06/italys-dpa-fines-glovo-owned-foodinho-3m-orders-changes-to-algorithmic-management-of-riders/
Questions
What is algorithmic management?
Algorithmic management is the use of software, not human supervisors, to assign, monitor, score and discipline workers. The system decides who gets work, how they are rated, and when their access to earning is cut off.
Has algorithmic management been ruled illegal?
Specific systems have. Italy's Garante ruled the rider-management algorithms of Deliveroo and Foodinho unlawful under the GDPR, and the Dutch regulator fined Uber for automated driver deactivation without human review. The pattern, not algorithms in general, is what got fined.
What does algorithmic management look like in the gig economy?
In the gig economy it means order-assignment, rating and deactivation decisions made by software. Deliveroo tracked riders' geolocation every 12 seconds; Uber automatically cut driver accounts on suspected fraud or low ratings, with no human assessment.
Sources
- Garante per la protezione dei dati personali, Rider: il Garante privacy sanziona Foodinho per 2,6 milioni di euro , 2021-07-05
- Garante per la protezione dei dati personali, Rider: sanzione di 5 milioni di euro a Foodinho , 2024-11-22
- Garante per la protezione dei dati personali, Rider: il Garante privacy sanziona Deliveroo Italy per 2,5 milioni di euro , 2021-08-02
- Autoriteit Persoonsgegevens (Dutch Data Protection Authority), Uber fined nearly 825 million euros for automated driver blocking , 2026-08-21
- TechCrunch (Reuters), Uber faces fine of nearly $1B over automated driver suspensions , 2026-08-23
This is analysis, not a verified outcome. It carries no verification badge and never will. The proof lives in the case files, where every figure is checked against the public record and the method is printed on the page.