You are the provider of an AI agent if you develop it, or have it developed, and put it on the market or into service under your own name. You are the deployer if you use it under your authority. Firms that build an agent on a third-party model for their own use are usually both, with both sets of duties.
What does the EU AI Act say about providers and deployers?
The EU AI Act (Regulation (EU) 2024/1689) assigns obligations by role. In substance:
- A provider develops an AI system, or has one developed, and places it on the EU market or puts it into service under its own name or trademark, whether paid or free. Putting into service includes supplying a system for the provider's own use.
- A deployer uses an AI system under its authority in a professional capacity.
For high-risk systems, providers carry the heavier load: risk management (Article 9), data and data governance (Article 10), technical documentation, automatic logging capability (Article 12), transparency and instructions to deployers (Article 13), design for human oversight (Article 14), accuracy, robustness and cybersecurity (Article 15), a quality management system, conformity assessment, registration and post-market monitoring.
Deployers of high-risk systems have duties under Article 26: use the system according to its instructions, assign human oversight to competent people, make sure input data under their control is relevant, monitor operation and report risks and serious incidents, keep the logs under their control, and inform affected workers and, for certain decisions, the people concerned. Some deployers must carry out a fundamental rights impact assessment under Article 27, including those using AI for creditworthiness assessment or life and health insurance pricing.
Both roles share duties on AI literacy (Article 4), and Article 50 splits transparency duties between them.
Who does this apply to?
Anyone placing AI systems on the EU market or using them in the EU, and providers and deployers outside the EU whose systems' output is used in the EU. The full obligations above attach to high-risk systems; other systems carry lighter duties such as transparency. For the role definitions on their own, see what a provider is under the EU AI Act and what a deployer is.
How do you tell which role you have for an agent?
Ask three questions about each agent.
Who built it? If your team assembled the agent, writing its brief, choosing its sources and connecting its tools, on top of a general-purpose model from a model provider, you have developed an AI system. The model provider is the provider of the general-purpose model; you are the provider of the agent.
Whose name is on it? If you bought an agent and run it under the vendor's name, the vendor is the provider and you are the deployer. If you put your own name or brand on a high-risk agent that someone else built, you can become its provider.
What have you changed? If you substantially modify a high-risk agent, or change an agent's intended purpose so that it becomes high-risk, you can take on the provider's obligations. Our piece on substantial modification covers this risk.
Most regulated firms end up in one of three positions. A bank that buys a vendor's credit assessment agent and runs it as supplied is a deployer. A bank that builds its own credit assessment agent on a third-party model and uses it internally is provider and deployer. A bank that buys an agent, then rewrites its criteria and extends it to a new high-risk use, may have become a provider without intending to.
What do the obligations look like side by side?
Take an illustrative case. An insurer builds an agent that recommends premiums for life cover, using a model from a model provider. Life insurance pricing is listed in Annex III, so the agent is high-risk. As provider, the insurer needs a risk management system, data governance, technical documentation, logging, instructions for its own users, oversight design, accuracy testing, a quality management system and conformity assessment. As deployer, it needs trained human overseers, monitoring, log retention and a fundamental rights impact assessment.
| Requirement | Control | Evidence it leaves |
|---|---|---|
| Provider: risk management (Art. 9) | Risk process per agent and capability through its life | Risk management file |
| Provider: data governance (Art. 10) | Sources and test data assessed for relevance and quality | Data governance records |
| Provider: logging (Art. 12) | Agent records every action and its basis automatically | Append-only action log |
| Provider: instructions (Art. 13) | Instructions for use stating purpose, limits and oversight needs | Instructions document |
| Provider: oversight design (Art. 14) | Held actions, stop controls, reviewer views | Design specification |
| Provider: accuracy and cybersecurity (Art. 15) | Testing on real cases; injection defences | Test and verification reports |
| Deployer: use per instructions (Art. 26) | Agent configured within the provider's stated limits | Configuration record |
| Deployer: human oversight (Art. 26) | Named, trained overseers with authority | Oversight assignments; training records |
| Deployer: monitoring and logs (Art. 26) | Operation monitored; logs kept | Monitoring reports; retained logs |
| Deployer: impact assessment (Art. 27) | Fundamental rights assessment before use where required | Signed assessment |
For deployer duties in detail, see Article 26 deployer obligations when you run an agent.
What will a supervisor or market surveillance authority ask?
- For each agent, are you the provider, the deployer or both, and why?
- If you built the agent, where is its technical documentation?
- If you bought it, what changes have you made, and could they make you its provider?
- Who are the named human overseers, and what training have they had?
- Which agents are high-risk, and on what Annex III ground?
What does the role split not settle?
It does not decide whether an agent is high-risk; that turns on its use. It does not settle duties under GDPR, sector rules or DORA, which attach to controllers, firms and financial entities on their own terms. And roles are assessed per system, so one firm can be provider of one agent and deployer of another. For the full regulatory map, see our guide to AI agent regulation mapped to controls.
This is an explanation of the rules, not legal advice; confirm how they apply to you with counsel.
Questions people ask
Is a model provider the provider of my agent?
Not usually. It provides the general-purpose model, with its own obligations. The agent built on top is a separate AI system, and whoever develops it and puts it into service is its provider.
Can a firm be provider and deployer of the same agent?
Yes. Building an agent and using it internally makes a firm both. It then carries provider duties for the system and deployer duties for its use.
Do deployers of agents that are not high-risk have obligations?
Fewer, but some. AI literacy applies, and Article 50 transparency duties may apply depending on what the agent does, such as interacting with people or generating certain content.