Delegated authority
Authority Register
Autonomy is only safe if somebody wrote down what each agent and each person is allowed to decide. This register is that document, and it is enforced: the decision route resolves a grant before it writes anything, and refuses when nothing covers the action.
Health
What the estate is currently allowed to do
Grants in force, who holds them, how often they are used and how often they are exceeded.
RD4-AUTH-H1
Director, Laboratory Operations · Research & Development · Lab Operations
Anything the agent grant refuses, plus any override of an agent outcome. An override must state what the agent concluded and why it is being set aside.
- Cannot approve a unit the same person raised.
- Override requires a written rationale citing the policy clause relied on.
- Expiry is enforced, not advisory: an expired grant refuses.
11 grants
Select any row to see the conditions, the delegation chain and how it has been used
| Reference | Holder | Decision type | Cap | Jurisdictions | Delegated from | Expires | Uses 30d | Status |
|---|---|---|---|---|---|---|---|---|
| RD6-AUTH-H1 Prototype & Pilot | Director, Product Development accountable-owner | Approve, override and release held work on the prototype & pilot lines | $2,000k | US UK DE SG | Research & Development leadership | 2026-11-05 | 259 | Active |
| RD4-AUTH-H1 Lab Operations | Director, Laboratory Operations accountable-owner | Approve, override and release held work on the lab operations lines | $25,000k | US UK DE SG AU FR | Research & Development leadership | 2027-02-06 | 24 | Active |
| RD8-AUTH-H1 Regulatory Science | VP Regulatory Science accountable-owner | Approve, override and release held work on the regulatory science lines | $2,000k | US UK DE SG AU FR BR | Research & Development leadership | 2026-10-10 | 98 | Active |
| RD11-AUTH-H1 Portfolio Analytics | Director, R&D Analytics accountable-owner | Approve, override and release held work on the portfolio analytics lines | $2,000k | US UK DE SG AU FR | Research & Development leadership | 2026-11-28 | 134 | Active |
| RD2-AUTH-H1 Innovation Portfolio | VP Innovation accountable-owner | Approve, override and release held work on the innovation portfolio lines | $250k | US UK DE SG AU FR | Research & Development leadership | 2026-10-28 | 29 | Active |
| RD7-AUTH-H1 IP Creation & Capture | Director, IP Capture accountable-owner | Approve, override and release held work on the ip creation & capture lines | $2,000k | US UK DE SG AU FR BR | Research & Development leadership | 2027-03-03 | 230 | Active |
| RD1-AUTH-H1 Discovery Research | Director, Discovery Research accountable-owner | Approve, override and release held work on the discovery research lines | $2,000k | US UK DE SG AU | Research & Development leadership | 2027-01-21 | 92 | Active |
| RD3-AUTH-H1 Experiment Design | Director, Research Methods accountable-owner | Approve, override and release held work on the experiment design lines | $250k | US UK DE | Research & Development leadership | 2027-02-24 | 22 | Active |
| RD5-AUTH-H1 Data Science & Modeling | Director, Data Science accountable-owner | Approve, override and release held work on the data science & modeling lines | $2,000k | US UK DE SG AU FR | Research & Development leadership | 2026-09-26 | 52 | Active |
| RD9-AUTH-H1 External Collaboration | Director, External Innovation accountable-owner | Approve, override and release held work on the external collaboration lines | $250k | US UK DE SG | Research & Development leadership | 2027-04-04 | 52 | Active |
| RD10-AUTH-H1 Technology Scouting | Director, Technology Scouting accountable-owner | Approve, override and release held work on the technology scouting lines | $2,000k | US UK DE SG AU | Research & Development leadership | 2026-10-06 | 99 | Active |
How the register is enforced
Not a wall chart
When a decision is recorded, the route resolves a grant first: the right function, the right holder kind, an active status, a cap at or above the value at stake, and the jurisdiction of the work. If no grant covers all five, the route returns a refusal and writes a refusal record into the evidence spine. The decision does not happen.
Refusals are evidence in their own right. A cap breach that was stopped is worth more to an examiner than a clean run, because it shows the control fires.
Expiry is a real state, not a label. An expired grant authorizes nothing, which is why the sixty-day expiry tile matters more than the active count.
What is still missing
Say it before an examiner does
The holder identity on a grant is a name, not a cryptographic identity. Nothing here proves the actor was the actor. Signed agent identity and attestation are separate work and are not built.
The persona a user is browsing under is chosen in the interface and is not yet an identity the server trusts. The register is enforced against the value and the function; it is not yet enforced against a verified caller.
Delegation is one hop deep. A real board delegation chain runs several levels, with sub-delegation rules and counter-signatures. That depth is modeled as a single delegated-from field here.
Actions
What is waiting on a person
A grant is a written permission with an expiry, a cap and a named delegator. Nothing below renews itself.
Renew or let a grant lapse
Expires within ninety days. On expiry the holder falls back to asking.
Explain a grant that was exceeded
148 attempts beyond the cap in the last thirty days across these grants.
Withdraw a permission nobody uses
Granted and never exercised. Unused authority is still authority.
Nothing else is pending
Every other grant is inside its cap and inside its term.
Operations
What this desk is allowed to start
A surface that only reports is not operable. This is the work this page can set in motion, and the bound it runs into.
Trigger and bound
This desk can grant, narrow, extend or revoke a permission, and every one of those is written to the spine with the person who did it. It cannot let a holder act above the materiality cap on the grant: an attempt above the cap is refused at the point of decision and recorded as a breach rather than allowed and flagged afterwards.
Live observability
What the record shows right now
Grants split almost evenly between agents and people. 352,403 uses in the last thirty days produced 148 attempts above cap.
Current distribution
309 grants
Is policy and strategy coming to fruition
Whether the written intent is holding here
148 cap breaches against 352,403 uses — 0% of exercised authority went past its bound.
Not holding on the record
The policy position is that no actor, agent or human, may commit the company past a stated value without a named delegator standing behind it. The record shows 309 active grants and 148 attempts above cap in thirty days, every one refused at the point of decision rather than allowed and reported later. That is the intent working. The part that is not proven is the ceiling itself: the caps were set by this team, not benchmarked against loss experience, so a grant being inside its cap says the rule held, not that the rule is right.