Virtue & Voltage

A few products are getting rewired right now. Full voltage resumes shortly.

Field notes

The escalation route: what “protected disclosure” actually means in practice

An escalation route is a defined path for raising a serious concern about an AI system that does not run through the person who has a reason to suppress it, with real protection for whoever raises it. Written like that, it sounds like a formality every organisation obviously has. In practice, most of what we find called an “escalation route” is an informal habit: raise it with your manager, who raises it with theirs, through the same chain that has every incentive to let a launch date survive intact.

The word doing the real work in that definition is “protected.”

A route that technically exists but offers no protection from being quietly deprioritised, reassigned, or managed out afterward is not a route…

We have seen organisations point to a policy document as evidence of an escalation route that had never been used.

What makes a route real, in our experience, is three things holding at once:

  1. it bypasses the person with a stake in the outcome,
  2. it is documented well enough to be pointed to after the fact, and
  3. there is a visible instance, even one, of it having been used.

The third one is the hardest to build and the easiest to fake with a policy nobody has tested. 

Without a working escalation route, an AI ethicist’s independence is just a claim.

Where this connects

A protected escalation route is one of the specific components the Organisation Toolkit's reporting-line and independence policy are built to make real, not just documented.

Next note →What EN 18274 will actually say, and when to expect it

© Virtue & Voltage. Quote a passage with a link back; please do not republish this note in full without asking.

← Back to Field notes