Step 1: the requirement and the scope of the policy
The NIS2 Directive in Article 21 paragraph 1 requires measures protecting network and information systems and their physical environment against incidents. Article 21 paragraph 2 lists among the measures human resources security, access control policies and asset management, and multi-factor authentication where appropriate. The access policy for server rooms and OT cabinets is therefore a document that follows from this list.
The scope of the policy is set by the list of places: server room, network cabinets, controller cabinets on the shop floor, control room. For each place write down the owner, the type of content and the way it is locked. Without the list the policy will be general, and the auditor will ask which doors it covers.
Step 2: roles instead of names
Write the policy in roles, not in names: server room administrator, automation engineer, visit supervisor, service company. Permissions in InnerWeb follow from roles with inheritance and a scope matrix, so the wording in the policy matches the system configuration. A change of person does not require a change of the document.
Roles are granted along a seniority ladder, and nobody can grant a role higher than their own. Write this rule into the policy as the rule for granting permissions: who grants them, to whom and up to which level. This is the answer to the question of who controls the administrators.
What the policy links with the evidence in the system
Step 3: risks assigned to places
Each place on the list has its own risks: fire, power failure, unauthorized connection, work near energized equipment. The catalogue of risks and protective measures can be assigned to a location, to a type of dangerous work and to a permit. As a result, access to an OT cabinet is not a general description but a list of measures that appear in the permit for work at that specific place.
In the policy it is enough to refer to the catalogue and to state the rule that the measures from the catalogue are part of every permit. The details remain in the system, where they can be changed with the author and the time of the change.
Step 4: the permit as a condition of entry for external companies
For contractors the policy states: entry to a server room or OT cabinet requires a permit or a planned visit with a supervisor. The electronic work permit defines the task, the place and the time window and passes through the preparation, main work and closing phases. The flow of issued permits lets the issuer and the contractor work on the same document.
At the gate, a planned visit replaces the one-day permit requirement, but trainings, qualifications and an active account still decide on entry. The visit has a supervisor, and the system calls the supervisor once, at the first entry. These three sentences are worth moving into the policy almost unchanged.
Step 5: authentication and administrative accounts
For administrative accounts the policy should require multi-factor authentication. A code sent by e-mail or TOTP is enabled per installation, with a trusted device period. Write into the policy for which roles it is mandatory and how long a device remains trusted. The directive refers to this measure where appropriate, so justify the choice with a risk analysis.
Step 6: evidence that the system produces
The policy must name the evidence: the entry and exit log, the permits, the trail of changes. Every scan at the security screen is an audit act and is written to the log. Changes of permits, the taxonomy of dangerous work, training results, automation rules and inspection routes are saved together with the author and the time. If the door is served by the client's access control system, InnerWeb receives card read events through a webhook and links them with qualifications, permits, attendance and working time records.
Write into the policy who reviews the evidence and how often. A periodic report with tables and a PDF attachment, sent daily, weekly or monthly, is a natural carrier of such a review.
Step 7: exceptions and failures
The policy must foresee what happens when the system does not work or a person has no permission and entry is necessary. Describe an emergency procedure with a supervisor and a later entry in the log. Remember places without network access: when the client's network blocks all outbound traffic from the application host, alarm webhooks are written to a durable queue collected by an external poller, so the alarm is not lost.
Step 8: review and audit trail
The policy has a review date and an owner. The audit trail answers the question who changed the data, and entities written only by the system are deliberately excluded from it, because their history is itself source data. The log stores local time of the Europe/Warsaw zone from October 2026; entries from the winter seasons 2023/24 to 2025/26 are shifted by one hour. Write this into a note to the policy, so that the auditor does not treat it as an error.
Checklist
Video
Frequently asked questions
Can one policy cover the server room and OT cabinets?
Yes, if it describes the types of places and the measures separately. A server room and a cabinet on the shop floor have different exposure, so the skeleton is shared while the details differ.
Must the policy name a specific system?
No. It should describe the rule and the evidence. The system is the way of enforcing it and can change without changing the policy.
How often should the policy be reviewed?
At least once a year and after a change in the scope of places or roles. The review and its date are part of the evidence for the auditor.
Sources
Turn the policy into evidence
We will show how the wording of the access policy matches the configuration of roles, permits and the log in InnerWeb. Arrange a conversation with the InnerWeb team.
Let's talk