Ex-Ante Authorization

Ex-ante authorization is the property that authority for a governed action is established before the action takes effect. The verdict is rendered as a precondition of execution and recorded as a verifiable artifact at the moment the verdict is rendered. Ex-ante stands in contrast to ex-post controls such as monitoring, logging, and audit, which describe what happened after the action has already taken effect.

A FERZ-defined regulatory and architectural property, foundational to On the Impossibility of Observability-Based Authorization: A Formal Impossibility Result for Ex-Ante AI Governance (Meyman, 2026) and operationalized in The Authorization Artifact Test: Applying the Impossibility Result to Ex-Ante Regulatory Regimes (Meyman, 2026).

Authority must precede effect

An action that has taken effect cannot be retroactively authorized. A database write that has committed, a transaction that has settled, a regulatory filing that has been submitted, a model that has been retrained on new data: these are facts in the world. After-the-fact review can characterize them, log them, or report them, but it cannot establish that they had authority to occur. The cost asymmetry runs in one direction. An unauthorized action that takes effect cannot be retracted; a blocked action can be re-evaluated. Authority that arrives after the action arrives too late.

Ex-ante is distinct from ex-post

Observability is ex-post. Monitoring is ex-post. Audit logs are ex-post. Telemetry, traces, dashboards, post-hoc reviews: all describe what happened after the action has already taken effect. They are valuable as instrumentation. They are not authorization. Observability is not authorization. AI governance is an enforcement problem, not a monitoring problem, because authority must be established before the action, not reconstructed after. The ex-ante and ex-post regimes operate on opposite sides of the action: ex-ante before, ex-post after. They are complementary, not substitutable.

What an ex-ante regime requires

An ex-ante authorization regime has three structural requirements. First, authority must be evaluated before execution, against an encoded policy that can be applied per-action. Second, the verdict must be recorded as a verifiable artifact at the moment the authorization verdict is rendered. Third, the artifact must be independently verifiable, which means a regulator, auditor, or counterparty can check the verdict against the inputs and policy without trust in the executing system. The FERZ corpus refers to this third property as structural independence; the recorded artifact is the Proof-Carrying Decision. Without all three properties, the regime is ex-ante in name only.

What ex-ante authorization is not

Ex-ante authorization is not a coarse approval workflow. A pre-deployment review that approves a system as a whole is ex-ante for the system but ex-post for every individual action the system will take afterward. Each action still needs authority before it takes effect. Ex-ante authorization is not a policy declaration: writing down a rule does not enforce it, and a policy enforced only by monitoring is ex-post in operation. Ex-ante authorization is not a "best efforts" or "policy intent" framing, which is aspirational rather than verifiable. And it is not certification: a system can be certified compliant and still produce unauthorized actions if its individual decisions are not gated before execution.

Where it sits in the framework

Ex-ante authorization is the regulatory category that the FERZ runtime authorization framework operationalizes. Execution-time authorization is the specific architectural implementation: per-action evaluation immediately before execution, at runtime. The runtime authorization boundary is the structural location at which ex-ante authority is enforced. Proof-Carrying Decisions are the evidentiary artifacts that make ex-ante decisions independently verifiable. Deterministic authorization is what allows those artifacts to be replayed and checked. Together these properties define how the ex-ante requirement is met in practice rather than asserted in principle.

References

  1. Meyman, E. (2026). On the Impossibility of Observability-Based Authorization: A Formal Impossibility Result for Ex-Ante AI Governance, v1.3.2. FERZ, Inc. https://doi.org/10.5281/zenodo.19647542
  2. Meyman, E. (2026). The Authorization Artifact Test: Applying the Impossibility Result to Ex-Ante Regulatory Regimes, v1.0. FERZ, Inc. https://doi.org/10.5281/zenodo.20013582
  3. Meyman, E. (2026). Execution-Time Authorization for AI Agents: A Formal Framework for Deterministic Governance Boundaries, v1.0. FERZ, Inc. https://doi.org/10.5281/zenodo.18764562

The full FERZ corpus is available at /research/papers, on Zenodo, and on SSRN.

Related concepts

Frequently Asked Questions

What is ex-ante authorization?
Ex-ante authorization is the property that authority for a governed action is established before the action takes effect. The verdict is rendered as a precondition of execution and recorded as a verifiable artifact at the moment the authorization verdict is rendered.
How is ex-ante authorization different from monitoring or audit?
Monitoring and audit are ex-post: they describe what happened after the action has already taken effect. Ex-ante authorization is established before the action, not reconstructed after. The two regimes are complementary, not substitutable. An unauthorized action that takes effect cannot be retroactively authorized by being observed.
Why must the verifiable artifact be recorded at the moment the authorization verdict is rendered?
Because the artifact is the evidence that authority existed before the action. A reconstruction assembled later cannot establish ex-ante authority; it can only describe ex-post behavior. The artifact must be produced at the moment the verdict is rendered and must be independently verifiable without trust in the executing system.
Is pre-deployment certification a form of ex-ante authorization?
No. Pre-deployment certification is ex-ante for the system, but it is ex-post for every individual action the system will take after deployment. Each action still requires authority before it takes effect. Certification can establish that a system has been approved to operate; it cannot stand in for per-action authorization.
What is the relationship between ex-ante authorization and execution-time authorization?
Ex-ante authorization is the regulatory and architectural property: authority must be established before the action. Execution-time authorization is one specific implementation of ex-ante: per-action evaluation immediately before execution, at runtime. Ex-ante is the requirement; execution-time authorization is one way of meeting it.
Cite this page

FERZ, Inc. (2026). Ex-Ante Authorization: Definition and Architecture. https://ferz.ai/governance/concepts/ex-ante-authorization

BibTeX
@misc{ferz2026exante,
  author    = {{FERZ, Inc.}},
  title     = {Ex-Ante Authorization: Definition and Architecture},
  year      = {2026},
  month     = {May},
  publisher = {FERZ, Inc.},
  url       = {https://ferz.ai/governance/concepts/ex-ante-authorization}
}