DORA has applied since 17 January 2025. Most financial entities spent that first year building registers, policies and incident procedures. In year two, the supervisory question changes: can you show that your operational resilience works?
That matters to banks, insurers and investment firms, and just as much to the ICT suppliers serving them. This article looks at DORA resilience testing, the state of third-party oversight, and the evidence to have ready before a supervisor asks for it.
From documents to demonstrated resilience
Year one was a transition. Attention went to the basics: registers of ICT contractual arrangements, and the classification and reporting of major incidents. The Dutch AFM reports that the share of registers accepted by the EBA rose from 40% in 2025 to 94% in 2026.
The harder findings concern substance. According to the AFM, most firms detect disruptions adequately, but preventive controls lag, notably logical access management and patch and vulnerability management. It also received fewer incident notifications than expected (65 major incident reports in 2025) and urged firms to review how they detect, classify and report.
Commentary such as IBM‘s expects 2026 to be the first full supervisory cycle, with ECB attention on cybersecurity, third-party risk and threat-led testing. That is an informed reading, not a published timetable.
A register tells a supervisor what you depend on. It does not show that you could keep operating if a dependency failed, and that gap is where testing comes in.
What threat-led penetration testing involves
Article 25 requires every in-scope entity to run a risk-based testing programme: vulnerability scans, network assessments, code reviews and penetration tests. Independent parties must run them, systems supporting critical or important functions must be covered at least yearly, and remediation must be provable.
Article 26 adds threat-led penetration testing (TLPT) for entities selected on impact, financial stability, ICT risk profile and maturity. TLPT is an intelligence-led red-team exercise against live production systems, built on the TIBER-EU framework. It generally recurs every three years, and entities that use internal testers must hire external ones for every third test.
A standard penetration test checks a defined scope for weaknesses. TLPT emulates a realistic threat actor against critical functions and tests whether people, processes and detection respond. After each test in the Netherlands, a Test Summary and Remediation Plan go to the supervisor.
An illustrative scenario: the red team sends a spear-phishing email from a lookalike domain to a treasury operator and captures a session token through an adversary-in-the-middle proxy, bypassing MFA. The real test is what follows. Does the SOC notice the new inbox rule, the unusual OAuth consent or the login from a new location?
The findings matter more than the intrusion itself. A test that reaches a critical system undetected is useful evidence of a gap, provided the remediation plan is credible and tracked to closure.
Third-party oversight: the list exists, the work has begun
Early commentary asked which providers would be treated as critical. That question now has an answer. On 18 November 2025 the ESAs designated critical ICT third-party providers; the AFM counts nineteen, and IBM’s account names AWS EMEA, Microsoft Ireland Operations, Accenture and IBM. Designation rests on systemic importance, support for critical functions and substitutability. Direct ESA oversight follows.
Questions remain. The ESAs plan to reassess criticality with 2026 data, so the list may change. Harmonising supervision across authorities of differing powers and maturity is also unfinished.
For financial entities, nothing transfers. A provider’s designation does not relieve you of accountability for contracts, registers and exit plans. Suppliers outside the list should expect more due diligence, audit and testing requests.
Concentration deserves attention too. If several critical functions rely on one cloud or email platform, an outage there is a single point of failure, designated or not.
A readiness check, prioritised by impact
- Map critical functions to systems and suppliers. Link each function to its ICT assets and register entries.
- Keep test evidence. Scope, reports, remediation owners and closure dates. A report with open critical findings and no owner reads as a risk accepted by default.
- Prove detection. SOC alerts and logs showing how an intrusion would surface, including email-borne initial access.
- Rehearse incident reporting. Measure the time from detection to classification to notification.
- Fix access and patching. These are the weak points the AFM names.
- Review supplier contracts. Check audit rights, exit strategies and willingness to take part in tests.
Items 1 to 3 come first because they produce the evidence the rest depends on.
Conclusion
Policies remain necessary, but they now open the conversation instead of ending it. Ask yourself one question: if a supervisor requested evidence next week, what would you send? The gaps in that answer are your work plan.
DORA places ultimate responsibility for ICT risk on the management body, so the answer cannot stay inside the security team.
LibraCyber’s Board Training DORA course is built for management, to help them ask for and read this kind of evidence.