Module 2 — Protecting Personal Information and Privacy

1. Why privacy matters in the internet age

Privacy is the ability to make informed choices about how information about us is collected, used, shared, and retained. It is not simply about hiding secrets. A person may willingly share a photograph with friends while reasonably objecting to that photograph being used to identify them, profile their interests, or train an unrelated system. Privacy therefore involves context, purpose, consent, fairness, and meaningful control.

In the internet age, ordinary activities leave digital traces: searches, purchases, location histories, connected-device readings, messages, and photographs. These traces can be combined to infer habits, relationships, health concerns, or financial circumstances. Even when an individual fact seems harmless, a collection of facts may be revealing. Privacy decisions often affect other people too, including family members, coworkers, and children.

AI tools add another destination for our information. We may paste a personal email, upload a household budget, or describe a medical concern to obtain useful assistance. The benefit can be real, but the disclosure may be unnecessary. The central question is not “Is this AI useful?” alone; it is also “What information must I provide, who may receive it, and what control will I retain?”

2. Learning outcomes

After completing this module, you should be able to distinguish personal information from especially sensitive information; explain how AI services may handle prompts and uploads; locate and interpret key privacy-policy statements; apply data minimization and need-to-know principles; and respond appropriately to an accidental disclosure.

3. What is personal information (PI)?

Personal information (PI) is information relating to an identified or identifiable person. The term personally identifiable information (PII) is widely used in U.S. privacy and security practice. The European Union’s General Data Protection Regulation (GDPR) generally uses the term “personal data.” These terms overlap, but their exact legal definitions and scope depend on the applicable framework.

Identification may be direct, such as a full name paired with an account number, or indirect, such as a combination of location, occupation, age, and unusual circumstances. A pseudonym or a removed name does not necessarily make information anonymous. Context matters: a ZIP code alone may be ordinary, while a ZIP code combined with a birth date and workplace may become identifying.

Some information deserves heightened care because misuse could cause substantial harm. Examples include medical information, financial account credentials, biometric identifiers, precise location history, and information about children. The GDPR also recognizes special categories of personal data subject to additional protections. The matrix below is a practical teaching aid, not an exhaustive legal classification.

Framework context: NIST SP 800-122 and GDPR Article 4. Examples are illustrative, not a complete legal taxonomy.

4. AI privacy and personal-information disclosure policies

An AI provider’s privacy notice, terms of service, and product-specific data controls describe what information the service collects, why it uses it, how long it may retain it, and when it may disclose it to vendors, affiliates, or other recipients. Enterprise and consumer versions of the same product can have materially different terms. Do not assume that every chatbot has the same protections.

Before submitting personal information, look for clear answers to these questions: Is my content used to improve or train models? Can humans review some conversations? How long are prompts, files, and logs kept? Can I delete or export them? Are there controls for shared links and connected apps? Where can I report a privacy incident? A policy that is hard to find or understand is a reason to reduce what you share.

Privacy policies are not a substitute for sound judgment. Even where a provider promises safeguards, account compromise, an accidental share link, a connected third-party application, or an incorrect user setting may create exposure. Review the relevant notice again after significant product or feature changes.

5. Data minimization and the need-to-know principle

Data minimization means using only the information that is adequate, relevant, and necessary for a particular purpose. In practical terms: if an AI can draft a complaint letter with [BANK], [AMOUNT], and [DATE], it does not need your account number, address, or full statement. Remove unrelated pages from documents, avoid pasting entire email threads, and describe a situation rather than attaching an identifiable record when possible.

Need-to-know is a closely related access-control principle: information should be available only to people or systems that genuinely require it for an authorized task. Applied to AI, ask not just “Do I need to share this?” but “Does this particular AI tool, integration, or recipient need to see it?” A travel-planning assistant may need a destination and preferences, but not your passport number or the exact dates your house will be vacant.

For example, ask an AI to explain the meaning of a common lab term without uploading a named test report; request a generic school-email template without entering a child’s name and classroom; or summarize a contract using a sanitized excerpt rather than an entire confidential agreement. These are privacy-preserving substitutions, not guarantees of anonymity.

6. How AI or LLMs may process information

When you send a prompt to an AI service, the service generally receives the submitted text and any files, images, or audio you attach. It may process that material to generate a response and operate supporting systems such as abuse detection, logging, storage, and safety review. The details depend on the provider, service tier, settings, location, and applicable agreements.

An AI service may also connect to search engines, email, calendars, document stores, or external applications. In such cases, data may move beyond the original chat interface. A connected application may have its own permissions, privacy terms, and retention practices. Review access scopes before enabling integrations, and disconnect tools you no longer use.

Deleting a visible chat is not always the same as immediately deleting all retained copies. Some services may retain limited information for security, legal, or operational reasons as explained in their policies. Conversely, it would be inaccurate to assume that every prompt is automatically used to train a public model. The correct approach is to verify the specific provider’s current terms and settings.

7. Ten key privacy risks

• Oversharing in prompts — A request for help with a dispute may inadvertently disclose your home address, account number, or family circumstances.
• Sensitive document uploads — One bank statement or medical record can contain many identifying details, not just the fact you wanted analyzed.
• Reidentification — Removing a name may not prevent identification when location, dates, occupation, or other clues remain.
• Unexpected retention or review — Service logs or review practices may differ from what a user assumes; check the actual policy.
• Third-party integrations — A connected app may gain access to chats, documents, or account data under separate rules.
• Compromised accounts — An attacker who accesses your AI account may see past conversations, files, or integrations.
• Accidental public sharing — Conversation links, screenshots, or collaborative workspaces can expose material to unintended people.
• Children and household members — A child’s photo, school, routine, or location can be disclosed without their understanding or consent.
• Hidden document information — Tracked changes, comments, spreadsheet tabs, and image metadata may disclose details beyond visible text.
• AI-generated redisclosure — An AI-produced summary, letter, or email may repeat private facts from the original input.

8. Practical actions for protecting privacy

Begin with a simple pre-submission review: identify the purpose of the prompt, remove identifiers that are not required, and consider whether a fictional or generalized example will work. If the task involves another person’s information, consider their expectations, permission, and any organizational obligations.

Inspect attachments carefully. For documents, remove unrelated pages, comments, tracked edits, and hidden worksheet tabs. For images, consider recognizable faces, street signs, school uniforms, and embedded location data. A “redacted” screenshot should be checked to ensure sensitive content is actually obscured, not merely covered by an editable overlay.

Protect the account and its connections. Use a unique password or passkey, enable multifactor authentication where available, review shared links, and periodically inspect connected applications. Read privacy settings with special attention to content retention and use for model improvement. Prefer employer-approved tools for work information.

If you accidentally disclose sensitive information, stop sharing it further. Remove public links if possible, review connected access, change exposed credentials immediately, and contact the relevant provider, employer, bank, or other institution through an official channel. Do not assume deleting the chat fully reverses the disclosure.

9. Everyday privacy examples

These scenarios show how an ordinary person can obtain the benefits of AI while disclosing less information. The safest approach is usually to separate the general task from the identifying details needed later, outside the AI service.

9A. Examples in practice

• Banking letter — Ask: “Draft a polite letter disputing an incorrect fee at [BANK] on [DATE] for [AMOUNT].” Insert real details only in the private final document.
• Family vacation — Request a packing list for a seven-day trip without sharing your home address, exact travel dates, or that the house will be empty.
• Health question — Ask for general explanations of a medical term or questions to bring to a clinician, rather than uploading an identifiable record.
• School communication — Request a template for discussing homework concerns without supplying a child’s full name, school, or teacher.
• Photograph caption — Describe the image in words rather than uploading a recognizable photograph of a child.
• Job or workplace writing — Ask for a generic project-summary format rather than uploading confidential customer records or internal strategy documents.
• Household budgeting — Ask for a budgeting template using invented figures rather than uploading statements with account numbers.

10. End-of-module checklist

Use this checklist before sharing information with an AI service.

1. 1. I identify the personal information contained in a prompt before sending it.
2. 2. I replace unnecessary names, addresses, dates of birth, and account numbers with placeholders.
3. 3. I never submit passwords, one-time codes, recovery keys, or payment security codes.
4. 4. I review documents, images, comments, hidden tabs, and metadata before uploading.
5. 5. I check the AI provider’s privacy notice, retention practices, and training-use settings.
6. 6. I use a unique password or passkey and multifactor authentication for my AI account.
7. 7. I check whether a share link or workspace makes conversations visible to others.
8. 8. I review connected applications and revoke permissions I no longer need.
9. 9. I protect other people’s information, particularly children’s and family members’ data.
10. 10. I know how to respond if I accidentally disclose personal or sensitive information.

Further reading and authoritative sources

• United States — NIST Privacy Framework — A voluntary framework for identifying and managing privacy risk.
• United States — NIST SP 800-122, Guide to Protecting the Confidentiality of PII — Foundational explanation of personally identifiable information and safeguards.
• United States — FTC Consumer Advice: Protect Your Privacy Online — Consumer-oriented privacy practices.
• European Union — GDPR, Article 4: Definitions — Legal definition of personal data and related terms.
• European Union — GDPR, Article 5: Principles — Purpose limitation and data minimization principles.
• Japan — Personal Information Protection Commission (PPC) — Japanese privacy regulator and information on personal information protection.