Outcome#
A researcher can start without guessing about responsibilities, repositories, data ownership, storage, compute, AI services, reviewers, or exceptions.
Concept#
Technical instructions cannot resolve missing ownership and access decisions. Assigning every student the same route also wastes time and leaves role-specific risks unaddressed.
Worked Example#
The plan grants no access by implication and clearly separates training completion from authorization.
A correct example uses these decisions:
- Does passing the passport authorize system access? No. Access requires a separate owner-approved operational decision.
- What access should be requested? Only what the person needs, with an owner and review or end date.
Common Trap#
Treating a passed training mission as authorization to access a system or dataset.
If Blocked#
Do not grant broad access while waiting for a decision. Record the blocker and obtain the information owner, system owner, or lab owner decision. Use lab IT policy for escalation.
Useful references:
Understand Before Accepting AI Output#
An agent may format the packet but cannot authorize access, classify data, approve AI services, choose compute shares, or create policy exceptions.