A technology firm can hold a current service-organization report, a customer contract with an uptime target, and a product roadmap for artificial intelligence, and still be unprepared for the afternoon it becomes someone else's operational incident. For many software, cloud, and SaaS companies, that is the real resilience question. The firm is not only a user of vendors. It is a third party to banks, health systems, utilities, agencies, and other businesses that have already defined the firm's platform as a critical service.
That inversion changes the work. Operational resilience is not only the firm's ability to keep its own operations running. It is the firm's ability to keep the service its customers treat as essential available, recoverable, and understandable when a subprocessor, a region, or a model fails. Third-party risk runs in both directions: the firm must manage its own suppliers, and it must be able to show customers, and in some cases supervisors, that it can be managed as a supplier. AI governance is part of the same promise whenever a model sits in the product or in the pipeline that keeps the product running.
Why this matters
SaaS and other technology firms often grow up with a security program, a reliability program, and a vendor-management program that do not share a map. Customers, however, experience those programs as one question: if this platform fails, or if a model in the product fails, can I still perform my critical work, and will I be told in time? A community bank, a health system, or a public agency that has classified the platform as important will look through the firm's SOC 2 report, contract, and incident process for evidence of availability, recovery, and control, not only for a security narrative.
The firm's own concentration can be just as material as its customers' concentration. A single cloud region, a single identity provider, a single payments or messaging subprocessor, or a single model provider can interrupt many customers at once. That is a third-party risk to the firm and a systemic concern for the customers who share it. For a broader discussion of shared dependencies and of what to do when a critical provider cannot recover, see The Hidden Risk of Concentration and When a Critical Provider Cannot Recover.
SOC 2 is evidence, not a resilience program
SOC 2 examinations are performed against the AICPA's 2017 Trust Services Criteria for Security, Availability, Processing Integrity, Confidentiality, and Privacy (with revised points of focus issued in 2022). The official AICPA resource describes SOC 2 as a report on controls at a service organization relevant to those criteria. Security is typically in scope. Availability, processing integrity, confidentiality, and privacy are included when they are relevant to the service and to the examination. For a platform that customers treat as a critical service, the availability criteria are often the resilience conversation disguised as an attestation discussion.
A SOC 2 report can be strong evidence of how controls were designed and, in a Type 2 examination, how they operated over a period. It is not, by itself, proof that the firm can recover a critical service within a customer's impact tolerance, that subprocessors have been tested, or that a model can be withdrawn without breaking the product. Customers who are financial institutions, health systems, or public agencies will read the report alongside their own third-party and continuity requirements. The firm's job is to make the underlying program, not only the report, able to answer those questions.
That is why availability commitments in customer contracts, incident-notification terms, recovery objectives, and subprocessor disclosure should be consistent with the controls described in the examination and with the program the firm actually runs. A report, a contract, and an internal plan that tell three different stories will be found out during an incident, if not sooner. For a broader discussion of proving third-party resilience rather than stopping at due diligence, see Beyond Due Diligence: Proving Third-Party Resilience.
When the customer is a financial entity, DORA may be in the room
Regulation (EU) 2022/2554, the Digital Operational Resilience Act, has applied since 17 January 2025. It sets harmonized rules for in-scope financial entities on ICT risk management, incident reporting, operational resilience testing, and ICT third-party risk. It also creates an oversight framework for ICT third-party service providers designated as critical. A US technology firm that provides ICT services to in-scope EU financial entities may therefore be on the other side of those duties even if the firm itself is not a financial entity. If the firm is later designated a critical ICT third-party service provider, a further oversight regime applies. Designation and scope are legal and factual questions. They are not questions a firm should meet for the first time in a customer questionnaire.
Practical implications appear earlier than any designation. In-scope customers will ask for a register-quality description of the service, contractual terms on security, incident notification, audit and access, subcontracting, and service-level and recovery expectations, and evidence that the firm can support testing. A purely domestic vendor program, or a SOC 2 report written only around security, may not answer those asks. US financial institutions without an EU entity may still use DORA as a comparison, but the regulation should not be described as applying to them solely on that basis.
AI in the product is a customer-resilience issue
Many technology firms now ship AI in the product, use AI in support and operations, or depend on a general-purpose model provider. That work has official frameworks even where no single US "AI act" applies to the firm as a technology company.
The National Institute of Standards and Technology published the Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1, on January 26, 2023. It is voluntary, use-case agnostic, and organized around Govern, Map, Measure, and Manage. It describes trustworthy AI as valid and reliable, safe, secure and resilient, accountable and transparent, explainable and interpretable, privacy-enhanced, and fair, with harmful bias managed. For a platform company, "secure and resilient" and "accountable and transparent" are the characteristics customers will test when a model is in the path of a critical workflow.
ISO/IEC 42001:2023, published in December 2023, is the first international management-system standard for AI. It specifies requirements for establishing, implementing, maintaining, and continually improving an AI management system in organizations that provide or use AI systems. Certification is voluntary and is not performed by ISO. For a firm that both uses AI internally and provides AI-enabled services, the standard is a way to put those two roles under one set of policies, lifecycle controls, and responsibilities.
Firms that place or deploy AI in the European Union also need to account for Regulation (EU) 2024/1689, the EU Artificial Intelligence Act. The Act entered into force on 1 August 2024 and applies in phases. Prohibited practices and certain early obligations have applied since 2 February 2025, governance and general-purpose AI rules since 2 August 2025, and most remaining obligations since 2 August 2026. Some high-risk AI rules apply later. A SaaS firm can be a provider, a deployer, or both, depending on how the product is designed and used. That classification, and whether a use is high-risk, is a legal and factual question. The resilience questions are available regardless: what is the intended purpose of the model, what are its limits, who can change it, how is it monitored, how are customers informed, and what happens to the service if the model or its upstream provider must be withdrawn?
See also Connecting AI Governance and Resilience and From AI Principles to Accountability.
Where the three issues meet
The meeting point is a customer workflow the firm has allowed its customers to treat as dependable: a processing path, a system of record, a reporting feed, or an AI-supported decision. If a subprocessor fails, a cloud region is lost, or a model is withdrawn, the firm's customers still have to serve their own customers and, in regulated sectors, their own supervisors. The firm's incident is then an operational-resilience event in someone else's program.
That is why subprocessor management, concentration analysis, recovery testing, and AI lifecycle control should be part of one map, and why that map should be explainable to a customer due-diligence team. Dependency mapping is discussed more fully in Mapping Dependencies Before Disruption. Exercises should include the loss of a critical subprocessor or model provider, not only a primary-region outage. Contracts with subprocessors should be at least as strong, on notification, audit, and transition, as the promises the firm makes to its own customers. Product and model changes should have a path to degrade or disable AI features without taking the whole service down.
What organizations should do
Build a single inventory of the services customers treat as critical, the subprocessors and infrastructure those services depend on, and the models in the product or in the operating path. Use that inventory to drive vendor management, reliability work, customer due-diligence responses, and AI governance.
Treat SOC 2 as evidence of a program, not as the program. If customers rely on the platform's availability, make sure availability controls, recovery tests, and incident processes are actually in scope for the way the service is run, and that they match contract language.
If you support EU financial entities, obtain a current legal reading of DORA as it applies to your role as an ICT third-party service provider, and be prepared to support information-register, contractual, testing, and incident needs. Do not wait for a critical-provider designation to organize that evidence.
Govern product and operational AI with a documented intended purpose, monitoring, human oversight where the use requires it, customer communication, and a discontinuation or degradation plan. Use the NIST AI RMF and, where a management-system frame is useful, ISO/IEC 42001. If you place or deploy AI in the EU, obtain a current legal reading of Regulation (EU) 2024/1689. Do not claim conformity or certification you have not independently established.
Give company leadership, and the customer-facing teams that answer diligence requests, one picture of concentration, recovery evidence, and AI risk. See also From Plans to Proven Readiness and Operational Resilience: Where to Start.
How OpResONE can help
OpResONE helps technology firms strengthen operational resilience, third-party and subprocessor oversight, and AI governance as a single management problem, including the design of evidence that customers and, where relevant, supervisors can actually use. Our work is independent. We have no affiliation with the AICPA, the European Supervisory Authorities, NIST, or ISO, and this article is not an official interpretation of their standards or regulations.
We bring BRILLIANCE to RESILIENCE®.
Related resources
- Beyond Due Diligence: Proving Third-Party Resilience
- The Hidden Risk of Concentration
- When a Critical Provider Cannot Recover
- Mapping Dependencies Before Disruption
- Connecting AI Governance and Resilience
- From AI Principles to Accountability
- From Plans to Proven Readiness
- Operational Resilience: Where to Start
Request a Strategy Call
If you want a focused reading of how your platform's resilience, subprocessor oversight, and AI governance look to the customers who treat your service as critical, talk with OpResONE about where to start.
Sources
- AICPA & CIMA, SOC 2 — SOC for Service Organizations: Trust Services Criteria
- AICPA, 2017 Trust Services Criteria (With Revised Points of Focus – 2022)
- Regulation (EU) 2022/2554 (Digital Operational Resilience Act), EUR-Lex
- ESMA, Digital Operational Resilience Act (DORA)
- NIST, Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1 (January 26, 2023)
- ISO/IEC 42001:2023, Information technology — Artificial intelligence — Management system
- Regulation (EU) 2024/1689 (EU Artificial Intelligence Act), EUR-Lex
- European Commission, AI Act overview

