← Back to Resource Library
Privacy & Security

AI & Data Privacy Reference Guide

Essential guardrails, subscription realities, and a workplace anonymization standard for anyone using consumer AI tools at work.

The Golden Rule

If you wouldn't print it and leave it on a public coffee shop table, do not paste it into a consumer AI prompt.

$20/month subscriptions grant faster processing and higher usage caps — not enterprise-grade legal protection.

1. Consumer $20/mo vs. Enterprise Tiers

TierUsed for training?Human review?
ChatGPT Plus ($20)ON by defaultYes (sampled)
Gemini Advanced ($20)ON by defaultYes (sampled)
Claude Pro ($20)OFF by defaultRare (flagged)
Enterprise / APINEVERNo access

2. The 3 Operational Privacy Zones

Zone 1 — Consumer Free / Default
Model training ON, human review active. Safe only for generic brainstorming, public code, and general research.
Zone 2 — Consumer Opted-Out
Model training OFF (opted out manually in Settings). Safe for general work tasks that don't involve confidential metadata.
Zone 3 — Enterprise / Cloud API
Strict data isolation under corporate agreement. Mandatory for PII, proprietary code, and strategy.

3. 3-Point Pre-Prompt Audit

4. Practical Anonymization Exercises

Exercise 1 · Client Relations & HR Escalation
Unsafe

"Draft a warning letter to account manager John Davis at Apex Logistics. He missed the $450k Miller Corp account renewal deadline on Aug 12."

Safe

"Draft a performance warning letter to an [Account Role] at a [Industry Company]. They missed the [Contract Value] client renewal deadline on [Date]."

Exercise 2 · Financial Analysis & Strategy
Unsafe

"Analyze our Q3 margin drop. Revenue in North Texas fell from $3.2M to $2.1M due to 14% supplier price hikes from Acme Parts Corp."

Safe

"Analyze a Q3 margin drop scenario. Regional revenue dropped from [Value A] to [Value B] due to a [X%] supplier price increase. Provide recovery strategies."

Exercise 3 · Code Refactoring & System Security
Unsafe

"Optimize this Python script connected to database db.prod.internal:5432 using password AdminPass2026! to fetch customer SSNs."

Safe

"Optimize this generic Python SQL query targeting a PostgreSQL instance to fetch [Record_Type] efficiently using parameterized inputs."