Safety, Privacy, and Responsible Use of Claude
Lesson 1: What Not to Share With Claude – Privacy in Practice
Lesson Objectives
By the end of this lesson, students should be able to:
- Identify six categories of sensitive information to protect in Claude conversations
- Apply anonymization and substitution techniques to use Claude effectively without sharing private data
- Understand their organization's data policies as they relate to AI tool use
Lesson Content
The privacy question to ask before every paste.
Before pasting any content into Claude, ask yourself: "If this conversation were visible to a third party, would this content create a problem?" That question covers personal data, organizational confidentiality, client information, and regulatory exposure.
This is not about distrust. It is about professional discipline. AI tools – like email, messaging apps, and cloud documents – are part of your digital environment. The same data hygiene habits you apply to those tools should apply here.
Categories to protect:
1. Personally identifiable information (PII): Full names, email addresses, phone numbers, physical addresses, Social Security numbers, passport numbers, date of birth combined with other identifiers, financial account numbers. Avoid pasting documents containing PII unless you have confirmed your platform's data handling is appropriate.
2. Client and customer data: Any information that belongs to your organization's customers, clients, or partners – purchase history, correspondence, contact records, account details. Your organization's data handling policies govern this data, and using it in AI tools may require specific approval.
3. Protected health information (PHI): Any information about a patient's health, treatment, diagnosis, or care. PHI is subject to HIPAA (and equivalent regulations in other jurisdictions). Do not use Claude to process identifiable health information unless you are operating under a data processing agreement that covers this.
4. Employee and HR data: Performance reviews, disciplinary records, compensation details, medical accommodations, personal circumstances shared in a professional context.
5. Organizational confidential information: Unreleased product plans, financial projections, merger or acquisition discussions, proprietary processes, competitive strategy.
6. Legal and regulatory sensitive material: Privileged attorney-client communications, ongoing litigation details, regulatory filings under NDA.
The anonymization technique.
You can usually accomplish your goal with Claude without sharing the sensitive data itself. Anonymize it:
- Replace real names with [PERSON A], [PERSON B], or role descriptions ("the manager," "the client")
- Replace specific financial figures with representative numbers ("the contract is worth $X") unless the exact amount is necessary
- Replace company names with "Company A" or role descriptions ("our largest client")
- Replace identifiers with placeholders that preserve the relationship without the identity
Then ask Claude your question about the anonymized version. The analysis, advice, or drafting help is equally useful – and no real data has been shared.
Check your organization's AI policy.
Many organizations have issued policies on what data can be used with which AI tools. Before using Claude for work-related tasks involving client or employee data, verify that your organization's policy permits it. If you are uncertain, ask your IT, legal, or compliance team. "I wasn't sure what our policy was" is not a defense if a data exposure occurs.
Enroll to continue this lesson.
The preview above shows the lesson objectives and opening lesson content. Enroll to view the full lesson, complete the practice work, and take the lesson quiz.
