The roles

Article 3 defines the operators the Act regulates.

  • Provider (Article 3(3)): develops an AI system or a general-purpose AI model, or has one developed, and places it on the market or puts it into service under its own name or trademark, whether for payment or free of charge. Most of the Act's obligations sit here.
  • Deployer (Article 3(4)): uses an AI system under its authority, except in the course of a personal, non-professional activity. A bank using a vendor's credit model is a deployer.
  • Importer and distributor: bring a system made outside the EU onto the EU market, or make it available further down the supply chain. They must check that the provider has done its part.

The roles attach to a particular system, not to the organisation. The same firm can be the provider of the chatbot it built and the deployer of the HR screening tool it bought.

When a deployer becomes a provider

Article 25 sets out when a distributor, importer, deployer or other third party becomes the provider of a high-risk system and takes on the provider's obligations in full. It happens if the organisation:

  • puts its own name or trademark on a high-risk system already on the market;
  • makes a substantial modification to a high-risk system so that it remains high-risk; or
  • changes the intended purpose of a system, including a general-purpose one, so that it becomes high-risk.

The third is the trap. Pointing a general-purpose assistant at CV screening, or using a general model to score credit applicants, can turn a deployer into the provider of a high-risk system, with the conformity assessment, technical documentation and quality management system that come with it. Configuration and integration choices can do this; no new model has to be built. The original provider is then no longer treated as the provider of that system, but must cooperate and hand over the information needed.

General-purpose models

For general-purpose AI models the same logic applies at the model level. Fine-tuning or otherwise modifying a GPAI model makes the modifier its provider only if the modification is significant: the Commission's guidelines set the indicative test at more than a third of the original model's training compute. Most enterprise fine-tuning falls well below it.

What each role has to do

For a high-risk system, the provider designs and documents it to meet the requirements in Articles 9 to 15, runs a quality management system, completes the conformity assessment, registers the system, monitors it after it is placed on the market and reports serious incidents.

The deployer uses the system according to its instructions, assigns competent human oversight, monitors its operation, keeps logs for at least six months, informs workers before workplace use, tells people when a high-risk system is used to make decisions about them, and in some sectors carries out a fundamental rights impact assessment. A deployer that spots a serious incident must tell the provider immediately.

The two depend on each other. A deployer cannot complete its impact assessment without the provider's instructions for use, and a provider cannot monitor real-world performance without the deployer's cooperation, which is why the role split belongs in the contract.

What good looks like

  • Record the role for every system in the AI inventory, with the reasoning, and revisit it when the system's use changes.
  • Check intended purpose before reuse. Any new use of an existing AI tool in an Annex III area triggers a role and classification review before go-live.
  • Put the split in the contract: who holds the technical documentation, who monitors, who reports incidents, and what the provider must hand over.
  • Watch for drift into the provider role through white-labelling, substantial modification or repurposing. The register's unclear AI ownership and accountability is where that drift usually starts.

Frequently asked questions

What is the difference between a provider and a deployer under the EU AI Act? The provider develops the AI system, or has it developed, and places it on the market under its own name. The deployer uses it under its own authority in a professional capacity.

Can a deployer become a provider? Yes. Under Article 25, putting your name on a high-risk system, substantially modifying it, or changing a system's intended purpose so that it becomes high-risk makes you its provider.

Is a company that fine-tunes a model its provider? Only if the modification is significant. For general-purpose models the Commission's indicative test is a modification using more than a third of the original training compute.

Can one organisation be both? Yes. Roles attach to each system, so a firm can be the provider of systems it builds and the deployer of systems it buys.

Where it sits in the register

Inadequate AI vendor due diligence is how a deployer ends up without the information its role requires, and unclear AI ownership and accountability is how it drifts into the provider role unnoticed. Third-party model and vendor due diligence and the AI ownership and accountability framework are the controls. The wider supply-chain picture is in third-party risk.