Skip to main content
    Why ReignFor Chief Risk Officers

    The fastest-growing part of your model estate sits outside the guidance that governs the rest.

    SR 26-2 revised US model risk guidance in April 2026 and put generative and agentic AI outside its own scope. OSFI E-23 went the other way and wrote AI/ML into its definition of a model. Either way the question lands with you, and it lands without a cycle to answer it on.

    A single-tenant dedicated instance, in your own AWS or Azure account or in infrastructure iTmethods operates. Both available today; Google Cloud is planned for 2027. Air-gapped and sovereign are in development.

    What guidance covers and what it does not
    Your model estate In scope of SR 26-2 Statistical and quantitative Non-generative AI models validated, monitored, inventoried under a framework you already built and already defend Out of scope, by footnote Generative AI Agentic AI The part that is growing the guidance hands the question back to your own governance SR 26-2, 17 April 2026, footnote 3 OSFI E-23 goes the other way: AI/ML is inside its definition of a model
    The gap is not that the cycle is too slow. It is that the fastest-growing part of the estate sits outside the guidance that governs the rest of it.
    The decision you are actually making

    A gap that is arithmetic, rather than opinion.

    Model risk management was built around a population you could enumerate and a change you could schedule. Neither assumption survives an agent that invokes a model a thousand times a day against a prompt somebody edited on Tuesday.

    01The guidance narrowed while the estate widened
    SR 26-2 supersedes SR 11-7 and states in footnote 3 that generative and agentic AI models are not within its scope.

    It also removed the at-least-annual validation cadence and narrowed the definition of a model. The principles still apply to traditional and non-generative AI models; the part of your estate growing fastest is handed back to your own governance.

    02Canada moved in the opposite direction
    OSFI Guideline E-23 defines a model to include AI/ML methods, and it takes effect on 1 May 2027.

    It sets review frequency by risk rating rather than by calendar, requires monitoring standards with defined thresholds and escalation, and prescribes the fields a model inventory has to carry. If you operate on both sides of the border, you are managing to two different scopes.

    03A prompt revision is a model change
    Prompt edits, routing changes and policy edits alter what the model does. If they sit outside change control, the validated model and the running model diverge quietly.

    That divergence is not detectable from a validation report. It is detectable from a record of what was actually authorized, at the moment it was authorized.

    04Supervisors are asking for continuous, not periodic
    The European Supervisory Authorities’ joint statement on frontier AI models, 31 July 2026, describes monitoring processes moving from periodic to continuous.

    It is a supervisory statement rather than a regulation, and it binds indirectly through supervised entities. It is also the clearest signal available about which direction the next set of expectations comes from.

    The four components

    Governance, validation, monitoring, change control — at the rate AI actually moves.

    The four components of model risk management, and what each needs from the infrastructure underneath before it can be evidenced rather than asserted.

    Own it

    Governance that names an owner

    Every model in use has a named owner, an approved purpose and a documented boundary, or it does not have governance — it has a spreadsheet. At the boundary, a model that is not on your policy list is not reachable, so the inventory and the enforcement are the same object.

    The question under it. Who owns the model that answered last night?
    Available today How policy is applied →
    Detect it

    Validation against a moving target

    A model validated in March is not the model answering in June if the provider shipped in between. The record carries which model answered each call and when, so a change of model is a thing you can see rather than a thing you have to be told.

    The question under it. How would that change reach you?
    Available today How routing is decided →
    Watch it

    Monitoring between cycles

    An annual review tells you about the year. It does not tell you what a model did on a Tuesday, which is the granularity an exception gets argued at. A record per call is the shortest interval there is.

    The question under it. What is your shortest reliable interval today?
    Available today What the record carries →
    Prove it

    Change control that holds

    Requirements become checks that run against work as it moves, and a check that cannot decide is routed to a named person rather than resolved quietly in the system’s favour.

    The question under it. Who can change a prompt in production?
    Co-design development Reign Assurance →
    What carries it

    The operating record your framework already assumes exists.

    Which limits applied at the moment of each action, what was authorized, what the outcome was, and who decided each exception. That record is produced by Reign Gateway, and it is the part of this that exists today.

    Reign GatewayPolicy, routing and a record on every model call. Available today. Reign OpsThe substrate the agents run on, operated to your change control. Available today. Reign AssuranceRequirements to checks to evidence. Co-design development, not sold.

    The record is hash-linked as it is written and exported write-once into storage you hold, under your retention and your Object Lock. Whether that satisfies your framework is a second-line judgment about model risk, and the risk function is the only party positioned to make it.

    The edge of it

    What this does not settle for you.

    A record is an input to a model risk judgment. It is not the judgment, and nothing here is a compliance conclusion.

    We do not tell you whether you are compliant.

    Reign produces operating evidence. What it means for your model risk framework is a second-line determination, reached by the people who own that framework and defensible to your supervisor.

    SR 26-2 is not an AI mandate, and we will not sell it as one.

    It states that it does not set forth enforceable standards, and it places generative and agentic AI outside its scope. Anyone citing it as a reason to buy AI governance is misreading it.

    Automatic failover across permitted models is not built.

    When a model does not answer, the call is recorded as failed. It does not fall back to a model your policy has not permitted, because it does not fall back at all.

    Reign Assurance does not issue an opinion.

    It does not certify, does not attest, does not issue an audit opinion and does not provide independent assurance, in any tense. Reign prepares the evidence; the people who own the risk decide.

    Regulatory references on this page are to SR 26-2, OSFI E-23 and the ESAs’ July 2026 joint statement on frontier AI models. Each is cited for what it says, including where what it says is that something is out of scope.

    Before the scoping call

    The questions that arrive before the scoping call.

    Does SR 26-2 require any of this?
    No, and we would rather say so than imply otherwise. SR 26-2 states that it does not set forth enforceable standards, and footnote 3 places generative and agentic AI outside its scope entirely. The argument for a record is not that a supervisor mandates one; it is that the guidance hands the question back to your own governance and your governance needs something to point at.
    We are federally regulated in Canada. Does E-23 change the answer?
    It changes the scope. OSFI Guideline E-23 defines a model to include AI/ML methods, sets review frequency by risk rating rather than by calendar, and prescribes the fields your model inventory carries — including monitoring status and next review date. It takes effect on 1 May 2027.
    What is actually recorded on a model call?
    Who called, what the policy decided, which model answered, and when. Refusals are recorded with the rule that refused them, and a call that gets no answer is recorded as failed. The records are hash-linked to each other and exported write-once into storage you hold. What the record carries →
    Is Reign Assurance available?
    No. It is in co-design development with the control functions that will have to rely on it, it is not sold and it is not shipping. The Gateway record described on this page is the part that exists today. The status register →
    AWS Advanced Tier Services Partner SOC 2 Type II on Reign Ops
    AWS Advanced Tier Services Partner and Validated Managed Service Provider. Twenty-one years operating critical infrastructure for regulated enterprises.
    Next step

    Bring us the model inventory you have to defend.

    We will look at what exists between validation cycles today, which of it is generative or agentic and therefore outside SR 26-2, and what a supervisor would find if they asked what a model did on a Tuesday.