RESEARCH / VERIFIED REMEDIATION

A fix is not finished until the evidence says it worked.

Containment can stop an immediate event. Verified remediation asks a harder question: did the correction remove the weakness while preserving legitimate behaviour?

Deep Grey Matter · Updated September 21, 2026

DGM Intent is in beta testing. Beta access is subject to availability and suitability. Joining the waitlist does not activate protection.

This is DGM’s proposed evaluation method for authorised environments. It is a methodology, not a published customer result or a claim of universal protection.

Establish the original evidence

Record the authorised scenario, affected assets, observed activity and expected outcome before changing the environment.

Separate observed facts from inference
Preserve the relevant event sequence
State what the test cannot establish

Apply an approved correction

Use the least disruptive containment or repair that addresses the identified problem. Keep sensitive changes under the organisation’s control and record what changed.

Require appropriate approval
Limit the scope of the change
Retain a reversible path where practical

Retest both sides

Repeat the suspicious path to check that it is blocked, then run a benign regression test to confirm the legitimate workflow still functions.

Verify the original weakness
Check for alternate paths
Confirm legitimate behaviour remains available

Report remaining uncertainty

A successful retest is evidence about a defined scenario, not proof that every related risk has disappeared. Record residual risk, coverage limits and the next review trigger.

Name untested assumptions
Document residual exposure
Define when verification should be repeated

DGM INTENT / EARLY ACCESS

Help shape practical AI security.

We are speaking with businesses facing real application, network and agent-security problems.

Apply for early access