Book a demo
Topics · REGULATORY

Payments Model Risk DORA | Moring

The Digital Operational Resilience Act, Regulation (EU) 2022/2554, has been fully applicable since 17 January 2025 and governs ICT risk management, third-party technology oversight, and operational resilience at payment institutions and other EU financial entities. It does not name AI explicitly, but AI systems used in payments workflows are ICT systems under its definitions, so model risk management must be integrated with DORA rather than run as a separate track.

On this page
  1. Short answer
  2. How it works
  3. FAQs
  4. Related reading
  5. Sources

Short answer

The Digital Operational Resilience Act, Regulation (EU) 2022/2554, has been fully applicable since 17 January 2025 and governs ICT risk management, third-party technology oversight, and operational resilience at payment institutions and other EU financial entities. It does not name AI explicitly, but AI systems used in payments workflows are ICT systems under its definitions, so model risk management must be integrated with DORA rather than run as a separate track.

How it works

The Digital Operational Resilience Act, Regulation (EU) 2022/2554, became fully applicable on 17 January 2025 and is now the governing framework for ICT risk management, third-party technology oversight, and operational resilience at banks, insurance firms, investment firms, payment institutions, and other financial entities regulated in the European Union. The regulation was drafted before the generative AI wave, and it does not explicitly name artificial intelligence or large language models in its text.

Sources: evecore.com: https://evecore.com/blog/dora-ai-compliance

The absence of the words does not create an exemption. AI systems used in financial services workflows are ICT systems under DORA's definitions, so the ICT risk management, third-party oversight, incident classification, and resilience testing requirements apply to them as they do to any other technology. Payment institutions that silo AI governance away from DORA create real supervisory risk, because the model estate and the ICT estate are the same estate when models screen transactions and agents act on payments.

Sources: evecore.com: https://evecore.com/blog/dora-ai-compliance

DORA solves a structural problem in EU financial regulation. Before it, financial institutions managed the main categories of operational risk mainly through capital allocation, but digital risk does not respond to capital the way credit risk does. The regulation sets a unified standard for technological and operational resilience across EU member states, requiring a cultural shift toward data governance, testing, and cybersecurity practice to protect critical digital infrastructure, and payment infrastructure is among the most critical there is.

Sources: everbridge.com: https://www.everbridge.com/blog/understanding-dora-how-to-operationalize-digital-resilience/ bluemetrix.com: https://www.bluemetrix.com/dora-guide-key-insights-compliance-best-practices/

For payments model risk, the practical bridge is evidence infrastructure. DORA compliance depends on timely, reliable evidence for incident investigation and reporting, and database capabilities that support it include append-only audit storage enforced at the storage layer, per-agent identity attribution on every write, strong read consistency across nodes, storage-layer idempotency, support for rollback and compensating transactions, and enforced data residency. An AI agent that moves money should write to infrastructure designed to prove what it did.

Sources: cockroachlabs.com: https://www.cockroachlabs.com/blog/dora-database-requirements-ai-agents/

The AI Act alignment question arrives on a defined schedule. From 2 August 2026, the EU AI Act expands obligations for banks using AI, and financial institutions should identify where AI is used, classify systems according to risk, implement appropriate governance controls, and align AI management with existing DORA operational resilience requirements. The AI Act does not prohibit banks from using artificial intelligence; it introduces a risk-based framework requiring governance, documentation, human oversight, and cybersecurity controls, particularly for systems classified as high risk.

Sources: atmmarketplace.com: https://www.atmmarketplace.com/blogs/eu-ai-act-2026-what-banks-need-to-know-about-ai-governance-dora-and-cybersecurity/

Integration beats parallel tracks. Running model risk management and ICT resilience as separate programmes produces duplicated evidence, unowned intersections, and gaps exactly where agents operate, at the junction of model behaviour and operational continuity. A payment institution that treats its fraud models, screening agents, and payment automation as ICT systems with model-level governance inside meets both regimes with one control plane, one inventory, and one incident process.

Sources: evecore.com: https://evecore.com/blog/dora-ai-compliance optro.ai: https://optro.ai/blog/navigate-ai-governance-and-regulatory-compliance-finance

The operating conclusion for payments leaders is to inventory AI against DORA's ICT register, extend incident classification to model and agent failures, include AI dependencies in third-party oversight and resilience testing, and enforce data residency and audit storage for agent writes. Executed this way, DORA becomes the resilience backbone that model risk management plugs into, and the 2026 AI Act milestones become a classification exercise over an estate that is already evidenced.

Sources: cockroachlabs.com: https://www.cockroachlabs.com/blog/dora-database-requirements-ai-agents/ atmmarketplace.com: https://www.atmmarketplace.com/blogs/eu-ai-act-2026-what-banks-need-to-know-about-ai-governance-dora-and-cybersecurity/

Leadership structure determines whether the integration holds. An effective AI governance model starts with executive accountability, establishing C-suite ownership of core AI initiatives rather than delegating the estate to technology teams alone, and compliance discipline matters especially during periods of heightened stress, exactly when payment operations are most tempted to bypass controls. Payment institutions that place AI under a named executive owner, tied to the DORA-aligned ICT register and the incident process, avoid the parallel-track failure and produce the unified evidence both regimes expect. The cultural shift the regulation demands, toward data governance, testing, and cybersecurity practice, is a leadership deliverable before it is a technical one, and payment institutions that treat it that way keep their resilience posture intact when the models change underneath it.

Frequently asked questions

Does DORA apply to AI systems used in payments even though it does not name AI?

Yes. DORA was drafted before the generative AI wave and does not explicitly name AI or large language models, but AI systems used in financial services workflows are ICT systems under the regulation's definitions. The ICT risk management, third-party oversight, incident classification, and resilience testing requirements therefore apply to payment AI. Treating AI governance as a separate compliance track from DORA is described in supervisory commentary as a siloing that creates real risk.

Which entities are in scope of DORA for payments operations?

DORA governs ICT risk management, third-party technology oversight, and operational resilience at banks, insurance firms, investment firms, payment institutions, and a range of other financial entities regulated in the European Union. It became fully applicable on 17 January 2025 and sets a unified standard across EU member states. A payment institution operating models or agents in the Union is in scope regardless of where the model vendor or the group parent is headquartered.

How should a payment institution integrate model risk management with DORA?

Integrate rather than parallel-run: inventory AI systems against the ICT register, classify model and agent failures within the incident process, include AI dependencies in third-party oversight and resilience testing, and require audit-grade storage for agent actions. Supporting infrastructure capabilities include append-only audit storage, per-agent identity attribution on every write, strong consistency, idempotency, rollback and compensating transactions, and enforced data residency. One control plane, one inventory, and one incident process serve both regimes.

How do the EU AI Act milestones interact with DORA for payment AI?

From 2 August 2026 the EU AI Act expands obligations for banks using AI, and financial institutions are advised to identify where AI is used, classify systems by risk, implement appropriate governance controls, and align AI management with existing DORA operational resilience requirements. The two laws address different objects but overlap in practice, so the efficient path is to run AI classification over an estate already governed and evidenced under DORA, rather than building a second programme alongside it.

Sources

  1. optro.ai
  2. bluemetrix.com
  3. - federalreserve.gov
  4. - ithandbook.ffiec.gov
  5. - artificialintelligenceact.eu
  6. - eur-lex.europa.eu