Private preview. Not launched, not indexed. No analytics, no cookies, and nothing you enter is sent anywhere.
For teams: Does your AI need the cloud?
Run It Local
Protocols / P-05
P-05

Draw the red line

Write a one-page AI use policy: what stays in-house, what AI may do, and what it won't.

Fixes
Boundaries
Clear lines on what never leaves, and what AI won't do.
Time
A week, including review
Part-time, alongside normal work.
Difficulty
Moderate
Honest estimate for a 50–500 person company.

IndicationsUse this when

  • Employees already use AI tools without guidance.
  • Leadership hasn't decided what must stay in the building.

ContraindicationsDon’t, or be careful, when

  • Not legal advice. Have counsel review the final policy.

EquipmentWhat you need

  • A blank page
  • Your access map from P-03
  • The NIST AI Risk Management Framework, for shared vocabulary

ProcedureSteps

  1. List what must never be pasted into outside tools: employee personal information, pay, investigations, unreleased plans.
  2. Name which approved tools people may use, and for what.
  3. Decide which knowledge should live on hardware you control.
  4. Write one line that AI answers questions and drafts documents, and does not make decisions about people.
  5. Note that automated tools used to screen or score candidates can carry legal duties, such as bias audits and notice under NYC Local Law 144.
  6. Have counsel review, then share it with a short explanation.

ChecksHow you know it worked

  • The policy fits on one page.
  • Managers can explain the red line in a sentence.

HandoffWho owns it next

People Ops owns the policy, with a yearly review.

Sources

  1. Microsoft and LinkedIn, 2024 Work Trend Index (31,000 knowledge workers, Feb–Mar 2024)
  2. Cisco 2024 Data Privacy Benchmark Study (2,600+ organizations, 12 geographies)
  3. NIST AI Risk Management Framework
  4. NYC DCWP, Automated Employment Decision Tools FAQ (Local Law 144)