Add explicit authority revalidation state after material change to SAFE evidence

[Sellynet]

Summary

SAFE already preserves important evidence about identities, permissions, approvals, tool calls, configurations, timelines, and incident context.

I would like to suggest making one additional distinction explicit:

a previously granted authority may remain unexpired and technically in scope, while no longer being reliable after a material change in the system or its operating conditions.

This is distinct from whether authority originally existed, and distinct from whether execution stayed within the scope that authority granted.

Problem

For agentic systems, the conditions under which authority was originally granted may change over time.

Examples may include:

• model or model-version changes;
• tool or capability changes;
• configuration changes;
• new data sources or access paths;
• delegation changes;
• policy changes;
• environment or infrastructure changes;
• changes in risk classification or operating context.

In such cases, forensic reconstruction may need to establish not only:

What authority had previously been granted?

but also:

What evidence establishes that the previously granted authority was still valid after the system or its operating conditions materially changed?

Without this distinction, an incident record may show that an agent possessed credentials, permissions, or a prior approval while remaining unable to establish whether the assumptions supporting that authority still held at the time of action.

Possible evidence elements

A minimal evidence model could preserve or reference:

• authorization source;
• authorization baseline and scope;
• authorization version or effective period;
• material-change events since authorization;
• whether revalidation was required;
• revalidation decision and timestamp, if applicable;
• authorization standing at the time of action;
• evidence provenance and freshness supporting that determination;
• revocation or suspension state;
• resulting enforcement or containment action.

Conceptually:

prior authorization → material change → revalidation → current authorization standing → enforcement

Relationship to existing SAFE discussions

I see this as complementary to existing issues rather than replacing them.

• Add Temporal Authority and Self-Extension Denial to the SAFE Evidence Framework #1 addresses temporal authority, time-bounded authority, and self-extension denial.
• Define machine-verifiable approval-to-execution scope binding for SAFE evidence #13 addresses whether execution remained inside the scope that was approved.

The question raised here is different:

even if authority has not expired and execution remains inside its original scope, should a material change trigger a requirement to establish that the authorization remains valid under the new conditions?

That distinction may be relevant to incident analysis, near-miss reconstruction, and reusable control recommendations.

Question for the working group

Would SAFE benefit from representing an explicit authorization revalidation / current-standing state following material changes, so that incident evidence can distinguish:

1. authority existed previously;
2. authority had not expired or been revoked;
3. execution was within scope;
4. the conditions supporting that authority had or had not materially changed;
5. revalidation was or was not required;
6. the authority was or was not considered currently valid at the time of action?

I would be interested in feedback on whether this distinction is already captured elsewhere in SAFE or related work, or whether explicitly representing revalidation after material change would improve the evidence model.

---

Disclosure: I am the founder of Astrynn Holdings and am conducting Aegis Research on current standing, material change, revalidation, and governed authority in agentic AI systems. I am raising this as an open research and evidence-model question, not as a request to adopt a proprietary implementation.