---
title: Do LLM providers belong in the DORA register of information?
url: https://opencontroller.lyzr.com/blog/dora-register-of-information-and-llm-providers
published: 2026-11-18
updated: 2026-11-18
author: Lyzr
---

# Do LLM providers belong in the DORA register of information?

**Usually, yes. Under DORA, a paid model API or an agent platform is an ICT service, so the contract behind it belongs in your register of information. If you reach a model through a cloud provider or a SaaS agent platform, the model developer may instead appear as a subcontractor in the ICT supply chain, and that mapping matters most when the service supports a critical or important function.**

This is general information, not legal advice. Your compliance team and competent authority decide how your register is built.

## What does the DORA register of information actually require?

DORA (Regulation (EU) 2022/2554) has applied since 17 January 2025. Article 28(3) requires financial entities to maintain and update a register of information on all contractual arrangements for ICT services provided by ICT third-party service providers. It is kept at entity level and at sub-consolidated and consolidated levels.

The register must distinguish arrangements that cover ICT services supporting critical or important functions. Entities report information from it to their competent authority at least yearly, and the authority can request the full register.

The standard templates sit in Commission Implementing Regulation (EU) 2024/2956, published on 2 December 2024. They cover the entity, each provider, the contracts, which group entities use the service, the ICT supply chain including subcontractors, and assessments for critical or important functions such as substitutability and exit plans. First submissions ran in spring 2025, with dates set nationally (for example 4 April 2025 in Ireland and 15 April 2025 in Luxembourg).

DORA defines ICT services broadly: digital and data services provided through ICT systems to internal or external users on an ongoing basis. A model API that takes prompts and returns completions on a subscription or pay-as-you-go basis fits that definition.

## How do model APIs and agent platforms map to the register?

The question is never "do we use AI?" It is "who do we have the contract with, and who sits underneath them?"

| How you use the model | ICT third-party provider in the register | What to map in the supply chain |
|---|---|---|
| Direct API contract with a model provider | The model provider | Its own hosting subcontractors, where known |
| Model accessed through your cloud provider's model service | The cloud provider | The model developer, if it materially underpins the service |
| SaaS agent features inside a CRM, ERP or ITSM platform | The SaaS vendor | The model provider the vendor uses, where the function is critical or important |
| Enterprise chat or coding assistant workspace | The assistant vendor | Hosting and model subcontractors per the vendor's sub-processor list |
| Self-hosted open-weight model | Your infrastructure or hosting provider | Usually no model vendor; record the model source in your own inventory |
| Agent framework library (open source) | Usually none, as there is no service contract | Track it in software inventory, not the register |

Two points trip teams up. First, click-through terms accepted with a corporate card are still a contractual arrangement. Second, a SaaS vendor switching its underlying model can change your supply chain without changing your contract, so ask vendors to notify you.

## Where do shadow AI keys break the register?

The register is only as complete as your knowledge of contracts. AI creates arrangements that procurement never sees:

1. **Expensed API keys.** A team signs up for a model API, pays by card, and builds an agent on it.
2. **Keys inside cloud accounts.** Model access is switched on in a cloud subscription the central team does not watch.
3. **Agent features turned on in existing SaaS.** The contract exists, but nobody re-assessed whether the service now supports a critical or important function.
4. **Developer tools.** Coding assistants with their own billing, adopted team by team.

Each of these is a register gap and, if the agent touches a critical or important function, a gap in your exit and substitutability assessment too. The fix is discovery first, then classification. See [how to find shadow AI agents](/blog/how-to-find-shadow-ai-agents) and [finding leaked and unused LLM API keys](/blog/find-leaked-and-unused-llm-api-keys).

## How do you decide if an AI service supports a critical or important function?

Use the same test you use for any ICT service. Ask whether a failure of the agent or model would materially impair your financial performance, the continuity of your services, or your regulatory compliance. In practice, check:

- Does the agent act in a customer-facing or payment process, or only draft internal text?
- Can a human take over if the model provider is down, and how fast?
- Is there a tested fallback model or provider?
- Does the agent hold delegated access to core systems?

If the answers point to critical or important, you need the fuller template fields, including alternative providers and exit plans. A tested model fallback is useful evidence here. Our broader view of DORA for agents is in [DORA and AI agents in financial services](/blog/dora-and-ai-agents-in-financial-services).

To connect this to other frameworks, see [how to map AI agent controls to compliance frameworks](/blog/how-to-map-ai-agent-controls-to-compliance-frameworks).

## How Opencontroller does this

Opencontroller auto-discovers agents, models and API keys across AWS, Azure and Google Cloud, and connects to agent platforms such as Salesforce, SAP, Databricks and HubSpot plus Claude Enterprise and OpenAI Enterprise workspaces. That gives you a list of which agents depend on which providers. The LLM Gateway attributes cost per agent and per version, which supports the expense fields, and its routing and fallback give you a tested alternative path. This is evidence for your register, not a substitute for it. For how a control plane fits, see [what is an AI agent control plane](/blog/what-is-an-ai-agent-control-plane), or [book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**Is an LLM API an ICT service under DORA?**
In most cases yes. DORA defines ICT services as digital and data services provided through ICT systems on an ongoing basis, and a model API fits that description.

**If we use a model through our cloud provider, do we list the model developer?**
The cloud provider is your ICT third-party provider. The model developer may need to appear in the supply chain mapping where it materially underpins a service supporting a critical or important function.

**Do open-source agent frameworks go in the register?**
Usually not, because there is no contractual ICT service. Track them in your software inventory and in your agent registry instead.

## Key facts

DORA (Regulation (EU) 2022/2554) applies from 17 January 2025. Article 28(3) requires a register of information on all contractual arrangements for ICT services from ICT third-party providers, distinguishing those supporting critical or important functions. Templates: Commission Implementing Regulation (EU) 2024/2956 (2 Dec 2024), including ICT supply chain and subcontractors. Model APIs and agent platforms are generally ICT services; models reached via a cloud provider or SaaS vendor appear as supply-chain subcontractors. Shadow AI keys create register gaps. Opencontroller discovers agents, models and API keys across AWS, Azure, Google Cloud and attributes LLM cost per agent; it provides evidence for the register. Pricing starts from $250K per deployment per year.

## Related

- [How does DORA apply to AI agents in financial services?](https://opencontroller.lyzr.com/blog/dora-and-ai-agents-in-financial-services.md)
- [What should third-party risk teams ask SaaS vendors about their AI agents?](https://opencontroller.lyzr.com/blog/third-party-risk-questions-for-saas-ai-agents.md)
- [How do you find leaked or unused LLM API keys?](https://opencontroller.lyzr.com/blog/find-leaked-and-unused-llm-api-keys.md)
- [More on Evidence and frameworks](https://opencontroller.lyzr.com/blog/cluster/evidence-and-frameworks)
