Skip to main content
June 9, 2026Anthropic Launches Claude Fable 5 with Runtime Fallback Safeguards and Mandat...

OPERATIONAL FRAMEWORKS

The Disposition Protocol

Six requirements that convert an accuracy signal into a formal governance decision. The detection layer tells you something is wrong. The Disposition Protocol specifies what the organization is required to do about it.

For every AI system operating in production, does your organization have a written document specifying who receives an accuracy signal, what threshold triggers a formal review, who can pause the system, what re-authorization requires, who must be notified within what timeframe, and within what window a formal decision must be documented?

v1.0 · May 2026Sougata Roy, sougataroy.com

Free to read and cite with attribution to Sougata Roy and sougataroy.com. Do not republish, rebrand, or claim authorship of any framework, term, or model as your own.

Disposition Protocol: The Six Requirements

v1.0
Detection signal · Governance decision
01

The Accuracy Owner

One named person receives every accuracy signal and ...

VERIFY
02

The Trigger Threshold

A specific number makes review mandatory instead of ...

VERIFY
03

The Suspension Authority

One named person can pause the system while review r...

VERIFY
04

The Re-Authorization Standard

Restart requires documented evidence, not business p...

VERIFY
05

The Notification Clock

The right people are notified within a written timef...

VERIFY
06

The Decision Window

A formal hold, resume, or escalate decision is docum...

VERIFY

The organizational gap

The monitoring works. The question is whether the organization was designed to act on what it finds.

Overview

Authorization is an organizational decision, not a platform setting.

This section explains why identity, permissions, logging, and access reviews do not by themselves prove that an agent was authorized to operate.

Use this section to separate technical access from formal organizational authorization before reviewing the control model.

Why this framework exists

Most AI governance structures are built around detection. Dashboards. Monitoring thresholds. Audit logs. Alert routing. The detection layer tells the organization that something is happening. What most governance structures do not contain is the layer that specifies what detection formally requires. An URGENT email becomes a status update. The status update becomes a Thursday agenda item. The agenda item becomes an action item. The action item produces no decision. The system keeps running. The Disposition Protocol is the governance design that closes the gap between detection and decision. It requires six things, written before deployment and attached to every AI system in production. Without it, the detection layer produces information. It does not produce governance.

The governance question

For every AI system operating in production, does your organization have a written document specifying who receives an accuracy signal, what threshold triggers a formal review, who can pause the system, what re-authorization requires, who must be notified within what timeframe, and within what window a formal decision must be documented?

Editorial governance illustration showing a detection signal becoming status updates and agenda items until a formal disposition decision is required.

Detection to decision

Monitoring is not governance until it requires a decision

The detection layer can show that something is wrong. The disposition layer defines what the organization must do about it.

Governing scenario

Read the scenario as an authorization failure, not a configuration mistake.

The scenario shows why an agent can expose sensitive data even when the platform is behaving exactly as configured.

Use this section to test whether your own agent approvals name the agent's purpose, access scope, accountable owner, and stop authority.

Authorization is the artifact

Authorization is the artifact. AI governance is the written record of who authorized the system to act, what it was allowed to do, and who answers when it does the wrong thing.

Editorial governance timeline showing pre-deployment authorization handing off to post-deployment response after a performance threshold is crossed.

Lifecycle handoff

The protocol completes the governance lifecycle

Pre-deployment authorization defines what may go live. Post-deployment disposition defines what happens when monitoring crosses a threshold.

The Control Model

The six requirements

The Disposition Protocol requires six things, written before deployment and attached to every AI system in production.

Use these controls as the minimum authorization review before any agent enters production operation.

Editorial governance map showing the six Disposition Protocol requirements attached to one authorization record.

Six requirements

Six written requirements attach to the production record

Accuracy owner, trigger threshold, suspension authority, re-authorization standard, notification clock, and decision window must exist before deployment.

01

Requirement 1: The Accuracy Owner

One named person, identified by name and role in the authorization record before the system goes live, who is formally required to receive every accuracy signal and owns the decision about what happens next. Not a team. Not a shared inbox. One person. When the monitoring fires, this person receives it and is formally obligated to act.

The Accuracy Owner is distinct from the Consequence Owner, who holds overall accountability for the agent's behavior. An organization can have a named Consequence Owner and still have nobody whose specific formal responsibility is the accuracy signal. The Accuracy Owner closes that specific gap.

What the platform provides

Agent 365 and Microsoft Purview generate the accuracy signals and surface them through monitoring dashboards and alert routing. The platform routes information. It does not assign the human obligation to act on it.

What the organization must design

A written record naming the Accuracy Owner, attached to the authorization record, before the system goes live. If the Accuracy Owner changes, the authorization record is updated.

02

Requirement 2: The Trigger Threshold

A specific number, written before deployment, at which the Accuracy Owner is no longer managing a concern and is formally required to initiate a review. Not a feeling. Not when we are worried. A number.

The threshold is an accuracy rate, a sensitivity rate, a reversal rate, or an error rate, depending on the system. It is chosen before deployment through the same governance process that authorizes the system. An organization that has monitoring but no threshold has a dashboard that produces information. It does not have a governance trigger.

What the platform provides

Monitoring infrastructure can surface the metric. The governance trigger begins only when the organization has written the threshold into the authorization record.

What the organization must design

The threshold documented in the authorization record, alongside the metric it applies to, the measurement method, and the measurement cadence. If the threshold is crossed, the Accuracy Owner's obligation begins automatically.

03

Requirement 3: The Suspension Authority

One named person, different from the Accuracy Owner, who holds the formal authority to pause the system while a triggered review runs. Named before deployment. If identifying the person who can pause a production system currently requires a committee meeting, the problem is not the alert threshold. The suspension authority was never designed.

The Suspension Authority exists so the pause decision can be made while the review is live, before the organization has converted a signal into another unresolved meeting note.

What the platform provides

Copilot Studio, Agent 365, and Microsoft Foundry provide suspension and deactivation capabilities. The platform can execute the pause. It cannot decide that the pause is required.

What the organization must design

The name and role of the Suspension Authority, documented in the authorization record, with a written procedure for how the suspension is executed. The procedure must be executable without calling the original developer and without access to the original development environment.

04

Requirement 4: The Re-Authorization Standard

Before the system resumes after a triggered review, specific things must be demonstrated and documented. What those things are is decided before deployment, not assembled under business pressure while the system is idle and the case for resuming grows louder by the hour.

The re-authorization standard defines what the Accuracy Owner and Suspension Authority must be able to show before the system is restarted: evidence that the root cause has been identified, evidence that the root cause has been addressed, evidence that the fix has been validated, and a documented decision by the Suspension Authority that the system is cleared to resume.

What the platform provides

The platform can help record, monitor, and execute the restart. It does not define the evidentiary standard that makes restart defensible.

What the organization must design

The re-authorization standard written into the authorization record before deployment. An organization that has suspension capability but no re-authorization standard has a system that can be paused and cannot be restarted with documented confidence.

05

Requirement 5: The Notification Clock

When a trigger threshold is crossed and a formal review begins, specific people are notified within a specific timeframe. The legal team. The board risk committee. The regulatory contact. Whoever the organization has determined has a right to know. The notification is not a retrospective disclosure assembled after the incident closes. It is a required communication with a documented timestamp that starts when the threshold is crossed.

The clock prevents the organization from treating notice as a cleanup task after the facts are settled. The obligation begins when the threshold is crossed, because that is the moment the governance process begins.

What the platform provides

Alerting and case-management tools can send and preserve notification evidence. They cannot decide who has a right to know or how long the organization is allowed to wait.

What the organization must design

A notification list and a maximum notification window, both written into the authorization record before deployment. The clock starts at the moment the Accuracy Owner confirms the threshold has been crossed.

06

Requirement 6: The Decision Window

Once a threshold is crossed and a formal review begins, a decision must be reached and documented within a defined window. The decision is one of three: hold, resume, or escalate. Not a committee meeting scheduled for next Thursday. A named deadline, logged at the moment the threshold is crossed.

The Decision Window exists because a triggered review without a deadline is not a governance process. It is a governance conversation. The difference is that one produces a documented decision within a defined timeframe. The other produces a Thursday agenda item.

What the platform provides

Agent 365 and Microsoft Purview can timestamp the trigger event and track elapsed time between trigger and decision. The platform can surface the window and log the outcome. It cannot enforce a deadline or define what the default action is if the window closes without a decision.

What the organization must design

A maximum decision window, written into the authorization record before deployment. The window specifies the latest acceptable time between threshold crossing and formal documented decision. If the window closes without a documented decision, the default action - pause or escalate - executes automatically. That default must be specified before deployment, not improvised under pressure while the system is still running.

Editorial evidence illustration showing testing protocols, oversight team, formal procedures, trigger review, and a missing process finding.

Case evidence

The failure is often the missing procedure

The governance record must show what the organization was required to do when the model showed signs of failing itself.

Pre-deployment authorization

The Organizational Agent Controls define what must exist before an agent goes live.

Post-deployment response

The Disposition Protocol defines what must happen when monitoring shows a system has crossed a performance threshold.

Primary sources

Research basis

The framework is anchored in primary-source identity, authorization, and zero-trust guidance.

Use this section to trace the framework's claims back to published source material.

Editorial governance illustration showing an examiner-ready disposition record from monitoring signal through notification, documented decision, and re-authorization evidence.

Examiner-ready record

The answer comes from records, not recollection

When monitoring fires, the organization can show who acted, when, under what authority, and what decision was documented.

mass.gov

Source: Massachusetts Attorney General, "AG Campbell Announces $2.5 Million Settlement With Student Loan Lender For Unlawful Practices Through AI Use, Other Consumer Protection Violations," July 10, 2025.

Open primary source ->

Case illustration

On July 10, 2025, the Massachusetts Attorney General settled with Earnest Operations LLC for $2.5 million over AI underwriting governance failures. As part of the settlement, announced by Attorney General Andrea Joy Campbell, Earnest was required to build written testing protocols, a named internal algorithmic oversight team, and formal procedures specifying what the organization was required to do when its AI models showed a problem. The state did not find a failing model. It found an organization that had no formal process for what to do when the model showed signs of failing itself. The Disposition Protocol is that process, written before the regulator has to write it.

A documented answer from records

Every AI system in production has a named Accuracy Owner, a written trigger threshold, a named Suspension Authority, a documented re-authorization standard, a notification list with a maximum notification window, and a defined decision window. When the monitoring fires, the Accuracy Owner receives it, the threshold determines whether formal review begins, and if it does, the Suspension Authority makes the pause decision without a committee meeting. The notification goes out within the documented window. The decision - hold, resume, or escalate - is documented within the defined decision window. If the window closes without a decision, the default action executes as specified in the authorization record. The re-authorization standard governs the restart. The regulator who asks what the organization did when the monitoring flagged a problem receives a documented answer from records, not from recollection.

Connected Frameworks

Choose the adjacent control point.

Organizational Agent Controls define what must exist before an agent goes live. These connected frameworks extend that record into accountability, readiness, and chain authorization.

Use this section after you understand the controls and need the next framework in the governance lifecycle.

70-minute operating sprint

Apply this framework in one working session

Use this as a live governance exercise. Leave the session with named evidence, a visible gap, and a next owner rather than another discussion note.

Working session board

One pass through the framework. One evidence trail.

6

Steps

70

Minutes

1

Owner

Live

Decision

01

10 minutes

Name the Accuracy Owner

Pick one production AI system and identify the named person who receives accuracy signals.

Output

Written evidence ready for the next governance decision.

02

10 minutes

Write the Trigger Threshold

Document the metric, number, measurement method, and cadence that begins formal review.

Output

Written evidence ready for the next governance decision.

03

10 minutes

Name the Suspension Authority

Identify the person who can pause the system without a committee meeting.

Output

Written evidence ready for the next governance decision.

04

15 minutes

Write the Re-Authorization Standard

List the evidence required before the system can resume.

Output

Written evidence ready for the next governance decision.

05

15 minutes

Start the Notification Clock

Write the notification list and maximum notification window for the triggered review.

Output

Written evidence ready for the next governance decision.

06

10 minutes

Define the Decision Window

Write the maximum window between threshold crossing and formal documented decision, and specify the default action if the window closes without one.

Output

Written evidence ready for the next governance decision.