§ DORA · AI Act COMPARISON

Two regulations, one AI system in production. Where DORA and the AI Act conflict, overlap and leave gaps.

Five concrete cross-references between Regulation (EU) 2022/2554 (DORA) and Regulation (EU) 2024/1689 (AI Act) — extracted from primary text and mapped to specific articles.

Summary

A financial entity deploying high-risk AI is now in scope of two parallel EU regimes. DORA (Regulation (EU) 2022/2554, in force 17 January 2025) governs ICT risk for the financial sector. The AI Act (Regulation (EU) 2024/1689, high-risk obligations 2 August 2026 as written, provisionally 2 December 2027 under the Digital Omnibus) governs AI systems regardless of sector.

Where the two regulations meet, they neither replace nor cleanly complement each other. Fontvera has extracted three article-level overlaps and two structural gaps from the primary text. None of them disappear because one regulation came first; both apply, simultaneously, to the same AI system in production.

Who this applies to
Financial entities (credit institutions, investment firms, insurance undertakings, crypto-asset service providers under MiCA, payment institutions) deploying AI systems for credit scoring, fraud detection, algorithmic trading, robo-advice or ICT risk management.
Compliance deadline
AI Act Article 50 transparency: 2 August 2026. AI Act Annex III high-risk: 2 December 2027 under the Digital Omnibus deal of 7 May 2026, pending Official Journal. DORA: in force since 17 January 2025.
§ Key articles

What the law says

AI Act Article 9
Risk management system across the entire lifecycle of a high-risk AI system.
AI Act Article 15
Accuracy, robustness and cybersecurity for high-risk AI systems.
AI Act Article 19
Automatic event logging — minimum six-month retention.
AI Act Article 99(2)
Up to €15,000,000 or 3% of worldwide turnover for breach of provider obligations.
DORA Article 5
ICT risk management framework.
DORA Article 10
Detection and anomaly identification of ICT-related incidents.
DORA Article 12
Backup, recovery and restoration policies.
DORA Article 17
ICT-related incident reporting.
§ Detail

In depth

Scope: which systems fall under both

DORA applies to 20 categories of financial entity listed in Article 2 — credit institutions, investment firms, insurance and reinsurance undertakings, crypto-asset service providers under MiCA, payment institutions, e-money institutions, central counterparties, central securities depositories, trade repositories, securitisation repositories, account information service providers, crowdfunding service providers, and ICT third-party service providers serving any of the above. Source: Regulation (EU) 2022/2554 (DORA).

The AI Act applies to any AI system placed on the EU market regardless of sector. Financial entities are not exempt. High-risk classification is triggered under Annex III, point 5(b) (creditworthiness assessment and credit scoring) and Annex III, point 5(c) (life and health insurance risk assessment and pricing). Source: Regulation (EU) 2024/1689 (AI Act).

An AI system used for credit decisions inside a bank is therefore inside both. The bank is a DORA financial entity and an AI Act deployer. The AI vendor is an AI Act provider and, under DORA, an ICT third-party service provider.

The three article-level overlaps Fontvera has mapped

Both regulations require the same general control to be implemented under different legal bases. Where one cleared an audit, the other does not — the documentation has to satisfy both.

AI ActDORAWhat is duplicatedSeverity
Article 15Article 10Robustness, cybersecurity and the ability to detect anomalies. AI Act demands accuracy/robustness on the AI system; DORA demands detection of ICT incidents at the infrastructure layer.Medium
Article 18Article 12Technical documentation and backup/recovery. AI Act requires 10-year retention of technical docs (Annex IV); DORA requires backup/restore policies tested annually.Low
Article 19Article 17Event logging and incident records. AI Act mandates automatically generated logs retained ≥6 months; DORA requires structured ICT incident records and notification to competent authorities.Low

Treating these as redundant is the failure mode. The AI Act log retention floor (six months) does not satisfy DORA where national rules demand longer; DORA backup obligations do not satisfy AI Act technical documentation rules under Annex IV.

The two structural gaps neither regulation closes

Fontvera flags these as gaps because neither text resolves the boundary on its own:

Penalty mismatch — and why the higher number wins

DORA does not set EU-wide fine ceilings. Member States set administrative penalties, and DORA Article 35 adds periodic penalty payments on ICT third-party providers up to 1% of average daily worldwide turnover per day of non-compliance, capped at six months. Source: DORA Article 35.

The AI Act sets fixed EU-wide ceilings under Article 99:

For an AI Act high-risk credit scoring system that fails conformity assessment, the AI Act ceiling is the binding number. DORA penalties stack on top under national law.

Real numbers Fontvera tracks

Which obligations come first when both apply

  1. AI Act conformity assessment under Article 43 — must be complete before the system is placed on the market or put into service. Fixed gate. No DORA equivalent.
  2. DORA ICT risk framework under Article 5 — must be in place before deploying any ICT system, including AI. Continuous obligation, not a one-time gate.
  3. DORA threat-led penetration testing under Article 26 — required at least every three years for significant entities. Where it covers a high-risk AI system, the test outcomes also feed AI Act Article 9 risk management.
  4. AI Act post-market monitoring under Article 72 — runs in parallel with DORA Article 17 incident reporting. Reports go to different authorities (market surveillance vs financial supervisor) on different timelines.

Authorities and reporting lines

DORA reporting goes to the financial entity's competent authority — typically the national central bank, securities regulator, or prudential authority — and to the European Supervisory Authorities (ESMA, EBA, EIOPA) where applicable. AI Act reporting goes to market surveillance authorities designated under Article 70 of the AI Act, plus the European AI Office for cross-border coordination. These are different bodies with different reporting templates and different deadlines (DORA: tight initial notification within hours; AI Act Article 73: serious incidents within 15 days).

What to do before 2 August 2026

If the same AI system is in production now and you are inside DORA scope, the work is mostly mapping, not new build:

  1. Reconcile your DORA Article 5 ICT risk framework against AI Act Article 9 risk management. They are not the same shape — model drift, bias and performance metrics need explicit owners that DORA does not name.
  2. Lock the log retention rule in writing. Where ePrivacy, GDPR Article 5 (storage limitation), AI Act Article 19 (≥6 months), and national DORA implementing rules pull in different directions, the longest justified retention wins.
  3. Draft your AI Act technical documentation under Annex IV with DORA Article 28 third-party clauses already mapped. The vendor will not deliver this twice for free.
  4. Run incident reporting drills under both regimes simultaneously. The window between AI Act Article 73 (15 days) and DORA Article 19 (initial within hours, intermediate, final) is where most teams currently fail audit walkthroughs.

Run your free AI Act compliance diagnostic

Five minutes. No login. Returns the exact AI Act classification (Prohibited / High-Risk Annex III / High-Risk Annex I / Limited-Risk / Minimal-Risk) plus the article list that applies.

→ Run the AI Act diagnostic


Cross-regulatory data update

Auto-merged from the Fontvera archetype dataset on 2026-05-12. The sections below are extracted verbatim from the `obligations` and `obligation_crossrefs` tables; the page itself was last reviewed manually before this update.

Summary statistics

Overlaps: 3 · Conflicts: 0 · Gaps: 2

5 article-level crossrefs catalogued between AI Act and DORA from the Fontvera EU regulatory corpus. Article numbers are verbatim from the underlying obligation_crossrefs table; descriptions are extracted, not paraphrased.

All crossrefs between these regulations

Article (A)Article (B)TypeSeverityDescription
AI Act Art 19DORA Art 17overlaplow[entity affected: Financial institutions providing high-risk AI systems] Both regulations require financial institutions to maintain and record logs of system events and incidents, with DORA specifyin
AI Act Art 15DORA Art 10overlapmedium[entity affected: Financial institutions deploying high-risk AI systems] Both regulations mandate that systems (AI or ICT) be designed for robustness, cybersecurity, and the ability to detect anomalie
AI Act Art 18DORA Art 12overlaplow[entity affected: Financial institutions] Both regulations require financial institutions to maintain technical documentation and backup/recovery procedures as part of their compliance and risk manage
AI Act Art ?DORA Art ?gaphigh[entity affected: Financial entities using AI for critical functions] Neither regulation explicitly defines the integration of AI-specific model risk management (e.g., model drift, bias monitoring) in
AI Act Art ?DORA Art ?gapmedium[entity affected: Third-party AI providers serving financial entities] DORA focuses on financial entities' resilience, while AI Act focuses on providers; there is a gap in clear contractual obligation

Overlaps explained

No conflict-type crossrefs were catalogued for this pair, but the 3 overlaps below mean a single control can be designed to satisfy both regulations at once. Plan the controls jointly to avoid duplicate effort:

Which regulation takes precedence

EU law does not lay down a universal precedence rule between AI Act and DORA. In practice three resolution approaches apply: lex specialis (the more specific provision wins when both purport to govern the same conduct); regulator guidance (EDPB, EBA, ESMA and the AI Office have all issued joint readings on overlapping articles — check the most recent applicable opinion); and document the choice (when the regulations leave the call to the controller, the audit defence is your written reasoning, not the regulator's silence). Where the corpus surfaces a conflict rather than an overlap, treat that as an escalation path to legal — not a control-design question.

§ Action items

Practical steps

01
Map every DORA Article 5 ICT control to its AI Act Article 9 counterpart in writing. Where DORA does not name a model-specific control (drift, bias, performance), assign an owner under Article 9.
02
Lock log retention at the longest of: AI Act Article 19 (≥6 months), national DORA implementing law, ePrivacy or GDPR storage limitation. Document the legal basis.
03
Draft the Annex IV technical documentation jointly with the AI vendor and embed DORA Article 28 third-party clauses in the contract — including audit rights on the model itself.
04
Establish a single incident-classification taxonomy that maps to AI Act Article 73 (serious incidents, 15 days) and DORA Article 19 (initial / intermediate / final) so one event drives both reports.
05
If the AI system is in production today, brief the competent financial supervisor on AI Act exposure. They will be asked the same question by ESMA/EBA/EIOPA before August.
§ What Fontvera found

Documents in our corpus

eiopa EU Fetched 2026-04
Opinion on Artificial Intelligence governance and risk management
eurlex EU Fetched 2026-04
EUR-Lex: 32025R0454 (2025-03-07)
ai_office EU Fetched 2026-06
§ Cross-references

Related Fontvera intelligence

Need a cross-border briefing on this?
Search Fontvera ↵ Run the AI Act diagnostic
AI Act Article 50 transparency
50 days
until 2026-08-02, when Article 50 transparency obligations apply (unchanged). Annex III high-risk obligations move provisionally to 2 December 2027 under the Digital Omnibus agreement of 7 May 2026, pending formal adoption.
Preparing for 2 August 2026? Read the EU AI Act August 2026 deadline requirements checklist.