A common assumption in web data projects is that whoever runs the scraper is responsible for the personal data it collects. Under the GDPR, and under the CCPA in a different vocabulary, that is often wrong. Responsibility follows who decides why the data is collected and how it is used, not whose servers send the requests. The organization that makes those decisions is the controller (a business, in CCPA terms); a company that only collects or handles the data on its behalf and under its instructions is a processor (a service provider or contractor). The labels matter because each role carries different obligations, and a mismatch between the contract and what actually happens is a gap worth checking in any project that collects public personal data.
Controller: the one who decides why and how
Under the GDPR, the controller is the person or organization that determines the purposes and the means of processing personal data. Purpose is the why: measuring sentiment about a product, building a research corpus, monitoring prices. Means are the essential parts of the how: which sources, which categories of personal data, whose data and for how long.
The controller carries the main obligations: having a lawful basis, informing people where required, honoring their rights, keeping the data secure, limiting what is collected and how long it is kept. In a web data project, the controller is usually the company that wants the dataset and decides what it should contain, even when it never writes a line of scraper code.
Hypothetically, a retailer that asks for the names and review texts of people who reviewed competing products, to analyze complaints, has decided the purpose and the essential means. It is the controller of that dataset whether its own team or an outside firm does the collecting.
Processor: acting on someone else's instructions
A processor handles personal data on behalf of a controller and only according to its documented instructions. A scraping provider that collects the fields a client specifies, from the sources the client chooses, delivers them and does not use them for anything else is acting as a processor for that work.
The GDPR requires a contract between controller and processor that covers, among other things, acting only on instructions, confidentiality, security, the use of sub-processors, help with individuals' rights and the deletion or return of data at the end of the service. Processors also have some obligations of their own, such as security and keeping records of processing.
The role depends on facts, not on the label in the contract. A provider described as a processor that reuses client data to build its own datasets, or to train its own products, is deciding its own purposes for that use and is likely a controller for it.
Joint controllers and independent controllers
Two organizations can decide purposes and means together. A research consortium in which two universities jointly design a social media study and share the dataset would typically be joint controllers. The GDPR expects joint controllers to set out between them who handles which obligations, such as answering access requests, and to make the essence of that arrangement available to the people concerned.
Two organizations can also each be separate controllers of the same data. A provider that collects public data on its own initiative, builds a dataset and licenses it to many clients is usually the controller of the collection; each client that buys a copy and decides its own use becomes a controller of its copy. Buying a ready-made dataset therefore does not transfer responsibility to the seller.
Sub-processors: the chain behind the processor
A sub-processor is a company a processor engages to help with the work, such as a cloud host that stores the collected data. Under the GDPR, a processor needs the controller's prior authorization, specific or general, before engaging sub-processors, and must pass on equivalent data protection obligations. In practice, this means asking any provider for its list of sub-processors and how it announces changes to that list.
How the CCPA names the same ideas
The CCPA, as amended by the CPRA, uses different words for a similar split, and its thresholds and details differ from the GDPR's:
| CCPA term | Rough GDPR counterpart | What defines it |
|---|---|---|
| Business | Controller | Determines purposes and means and meets the law's thresholds |
| Service provider | Processor | Processes data on behalf of a business under a written contract that limits its use |
| Contractor | Processor | Receives data from a business for a business purpose under a similarly restrictive contract |
| Third party | Independent controller | Anyone else who receives the data, which can make the transfer a sale or sharing |
The practical point is the same as under the GDPR: without the required contract terms, a vendor that would otherwise be a service provider may be treated as a third party, and the disclosure may be treated as a sale or sharing with its own notice and opt-out duties. Some US states also have registration rules for data brokers, which can apply to companies that sell personal data about people they have no direct relationship with.
How the roles map onto common web data setups
- In-house collection for your own analysis: your organization is the controller; any cloud host is a processor.
- Custom collection by a provider to your specification: you are usually the controller and the provider a processor, provided it uses the data only for you.
- Buying an existing dataset: the seller is a controller of its collection; you become a controller of your copy and your use.
- A provider that collects, enriches and reuses data across clients: the provider is likely a controller for that reuse, alongside each client.
When evaluating any provider, ScrapingIsNotACrime included, ask which of these setups applies, in writing, and check that the contract matches. The overview of ethical web scraping and data privacy standards covers the questions about minimization and security that belong in the same review.
Common mistakes with data protection roles
- Assuming the provider carries the risk. A controller that outsources collection still answers for having a lawful basis and for honoring people's rights.
- Signing a processor contract that does not match reality. If the provider decides which personal data to collect, or reuses it, the contract describes a relationship that does not exist.
- Forgetting the end of the contract. Decide in advance whether data is returned or deleted, including copies in backups, and how deletion will be confirmed.
- Ignoring the people in the data. Whoever is controller needs a way to receive and answer access, deletion and objection requests, even for data taken from public pages.
- Treating roles as fixed. A project that starts as pure collection for one client can drift into reuse. Review the roles when the purpose or the data changes.
Where roles are unclear, or the data includes sensitive categories or people in several countries, have a lawyer familiar with the relevant jurisdictions review the setup and the contracts before collection starts. The myths about collecting public personal data under the GDPR and CCPA put these roles in the wider context of privacy rules.