The EU AI Act became the world’s first comprehensive AI regulation when it entered into force in August 2024. The obligations that matter most to financial institutions, those governing high-risk AI systems under Annex III, apply from 2 December 2027, after the Digital Omnibus deferred them from 2 August 2026. Transparency obligations under Article 50 already apply, from 2 August 2026.
If your organization deploys AI agents for credit scoring, KYC automation, or portfolio management, this article is for you.
Which Financial Services AI Is High-Risk?
Annex III lists the AI systems that are high-risk under Article 6(2). Three of its categories reach financial institutions directly:
- AI systems intended to be used to evaluate the creditworthiness of natural persons or establish their credit score (Annex III, point 5(b))
- AI systems intended to be used for risk assessment and pricing in relation to natural persons in the case of life and health insurance (point 5(c))
- AI systems intended to be used for the recruitment or selection of natural persons, to make decisions affecting work-related relationships, or to monitor and evaluate the performance and behaviour of persons in such relationships (point 4)
Point 5(b) has an exception: it does not cover “AI systems used for the purpose of detecting financial fraud”. An agent used to detect financial fraud is not high-risk under that point.
So an AI system used to assess the creditworthiness of natural persons or establish their credit score is high-risk under Annex III. For other finance uses, such as KYC automation or portfolio management, it depends on the use case: a system is high-risk under Article 6(2) only if it falls within an Annex III category.
Who Owes What: Providers and Deployers
The Act splits obligations between two roles. A provider develops an AI system, or has one developed, and places it on the market or puts it into service under its own name or trademark (Article 3(3)). A deployer uses an AI system under its authority (Article 3(4)). A financial institution that develops its own high-risk system and puts it into service under its own name is its provider. One that uses a vendor’s system is its deployer.
Provider obligations: Articles 9 to 15
Article 16(a) requires providers to ensure their high-risk AI systems comply with the requirements in Chapter III, Section 2 (Articles 8 to 15). These are obligations of the provider, not the deployer:
- Article 9, risk management system: “A risk management system shall be established, implemented, documented and maintained in relation to high-risk AI systems.”
- Article 10, data and data governance
- Article 11, technical documentation
- Article 12, record-keeping: high-risk AI systems “shall technically allow for the automatic recording of events (logs) over the lifetime of the system.”
- Article 13, transparency and provision of information to deployers
- Article 14, human oversight
- Article 15, accuracy, robustness and cybersecurity
Providers must also ensure the system undergoes the relevant conformity assessment procedure under Article 43 before it is placed on the market or put into service (Article 16(f)).
Deployer obligations: Article 26
Article 26 sets out the obligations of deployers of high-risk AI systems. Among them, deployers must:
- take appropriate technical and organisational measures to ensure they use the system in accordance with its instructions for use (Article 26(1))
- assign human oversight to natural persons who have the necessary competence, training and authority, and the necessary support (Article 26(2))
- to the extent they exercise control over the input data, ensure it is relevant and sufficiently representative (Article 26(4))
- monitor the system’s operation on the basis of its instructions for use and, where they have reason to consider it may present a risk, inform the provider or distributor and the market surveillance authority and suspend its use (Article 26(5))
- keep the logs the system automatically generates, to the extent the logs are under their control, for at least six months unless other law provides otherwise (Article 26(6))
- before putting a high-risk system into use at the workplace, inform workers’ representatives and the affected workers (Article 26(7))
- where an Annex III system makes, or assists in making, decisions related to natural persons, inform those persons that they are subject to the use of the high-risk AI system (Article 26(11))
Two provisions apply to financial institutions specifically. For deployers that are financial institutions subject to requirements regarding their internal governance under Union financial services law, the monitoring obligation in Article 26(5) is deemed fulfilled by complying with those internal governance rules (Article 26(5)), and the logs are maintained as part of the documentation kept under that law (Article 26(6)).
The Audit Evidence Problem
Here’s the compliance gap most financial institutions aren’t solving yet: producing audit evidence for agentic AI is fundamentally different from producing evidence for traditional model risk management.
A traditional model has inputs, outputs, and a defined decision boundary. You can validate it with holdout datasets, document its performance, and produce a static report.
An AI agent is different. It:
- Makes sequences of decisions in real time
- Calls external tools and APIs
- May behave differently on every run
- Can produce different outcomes for identical inputs due to LLM non-determinism
Traditional model risk management documentation (think SR 11-7) was not designed for this. The EU AI Act’s record-keeping rules apply to high-risk AI systems all the same: Article 12(1) requires them to allow the automatic recording of events over their lifetime, and Article 26(6) requires deployers to keep those logs.
What Needs to Happen Before 2 December 2027
Immediate (Now)
- Inventory all AI agents in production and classify their risk level
- Identify which are high-risk under the EU AI Act
- Assess your current logging and audit trail capabilities
Near-Term (Next 90 Days)
- Instrument high-risk agents with runtime logging that captures every action, policy evaluation, and decision point
- Assign human oversight of each high-risk system to people with the necessary competence, training and authority (Article 26(2))
- If you are the provider of a high-risk system, begin its technical documentation (Article 11)
Before 2 December 2027
- If you are the provider, complete the conformity assessment under Article 43 before the system is placed on the market or put into service (Article 16(f))
- If you are the deployer, monitor each high-risk system’s operation as Article 26(5) requires
- Ensure your audit trail can produce a complete evidence package on demand
How AgentGovern Addresses These Requirements
AgentGovern was built specifically for this compliance problem. Our platform:
- Instruments agents at runtime via two SDK calls and a callback handler passed to your agent, with no changes to your business logic
- Generates audit logs of every agent action passing through AgentGovern, with regulatory references
- Maps each action to EU AI Act provisions: Articles 9, 10, 12, 13 and 14
- Produces EU AI Act conformity reports from the captured evidence
The deadline moved to 2 December 2027, but the work did not get smaller. The organizations that will meet it are the ones instrumenting their agents now.