How to Protect Data Before ChatGPT: A Complete Guide

ChatGPT is one of the most widely used AI tools in the world, and many of the people who use it never check what happens to the data they put in.

Here is the uncomfortable truth: when you paste a document into ChatGPT, that data travels to OpenAI's servers, stays in your chat history until you delete it, and, depending on your plan and settings, may be used to train future AI models. Even deleted chats and Temporary Chats may be kept for up to 30 days under OpenAI's stated policy at the time of writing. If that document contains your Social Security number, your client's home address, your company's financial data, or any other personally identifiable information, that PII is now in OpenAI's hands.

This is not fear-mongering. OpenAI's own privacy policy states they collect and use personal information from ChatGPT conversations. And the FTC's guidance for businesses expects companies to protect the personal information they hold, including when service providers handle it.

The good news: you can use ChatGPT productively without exposing sensitive data. This guide walks you through exactly how — from understanding ChatGPT's data policies to configuring privacy settings to stripping PII from documents before they ever reach OpenAI's servers. For a broader look at PII protection across all AI tools, see our PII redaction before AI playbook.

What ChatGPT Actually Does With Your Data

Understanding ChatGPT's data practices is the foundation of protecting yourself. Here is what OpenAI's published privacy policy and help pages said at the time of writing. Check them again before you rely on any detail, because they change.

Data Collection

According to OpenAI's privacy policy, the company collects:

• Account information: Name, email, phone number, payment details
• User content: The text, images, and files you input into ChatGPT, plus the outputs generated
• Usage data: How you interact with the service, including conversation patterns and feature usage
• Device data: IP address, browser type, operating system

The critical item is user content — everything you type or upload. This means every contract clause, every personal detail, every financial figure, and every client name you input becomes data that OpenAI collects.

Data Retention

Chats stay in your history until you delete them. At the time of writing, OpenAI says deleted chats are scheduled for permanent deletion within 30 days, unless it must keep them for security or legal reasons. Temporary Chats are not used for training but may also be kept for up to 30 days for safety monitoring. Enterprise and Team plans come with stronger data controls and no training on business data by default.

Model Training

This is the most important point: on consumer plans, OpenAI may by default use your ChatGPT conversations to train and improve its models. Their privacy policy states they use personal information "to develop and improve our Services." You can opt out of model training through the data controls settings (Settings > Data Controls > Improve the model for everyone), but this is not enabled by default — you must actively disable it.

The March 2023 Data Incident

In March 2023, OpenAI confirmed a ChatGPT bug that exposed user data. According to OpenAI, some users could see titles from other users' chat history, and payment-related information (such as the last four digits of a card number and its expiration date) of about 1.2% of ChatGPT Plus subscribers who were active during a specific nine-hour window may have been visible to others. OpenAI took ChatGPT offline to fix the issue. Even well-resourced AI companies have incidents, which is why protecting data before upload matters.

Step-by-Step: How to Protect Your Data Before ChatGPT

Whether you are on the free plan or an enterprise subscriber, these steps protect your data before it reaches OpenAI's servers. The whole process adds a few minutes to your workflow.

Step 1: Configure ChatGPT's Data Controls

Before using ChatGPT for any sensitive work, configure your privacy settings:

1. Open ChatGPT and click your profile icon
2. Go to Settings > Data Controls
3. Toggle off "Improve the model for everyone" — this prevents your conversations from being used for model training
4. Consider using Temporary Chat mode for sensitive conversations — this provides an additional layer of data minimization

Important: Turning off model training does not stop retention. Your chats stay in your history until you delete them, and OpenAI's stated policy at the time of writing allows up to 30 days after deletion for safety and legal reasons. That is why PII redaction before upload still matters, even with the strictest settings.

Step 2: Audit Your Document for Sensitive Data

Before uploading any document to ChatGPT, do a quick mental scan:

• Does this document contain names of real people?
• Does it contain Social Security numbers, dates of birth, or government IDs?
• Does it contain home addresses, phone numbers, or email addresses?
• Does it contain financial data — bank accounts, salary figures, tax IDs?
• Does it contain medical information or health-related data?
• Does it contain proprietary business information — trade secrets, source code, internal strategies?

If the answer to any of these is yes, proceed to Step 3 before uploading.

Step 3: Redact PII With an Automated Tool

Upload the document to a free PII redaction tool first. Justee's PII redactor detects names, organizations, addresses, contact details, dates of birth, and government and financial ID numbers (30+ types, best-effort), and replaces each one with a numbered placeholder; amounts such as salaries are not redacted. The process typically takes a few minutes, runs on Justee's servers, and the original and the redacted copy are deleted once you download. Justee tells you how many items it found and removed, by type; open the redacted copy and check it against your original before you share it. Detection is best-effort and may not catch every identifier.

The result keeps the parts that matter for analysis, such as contract clauses, legal provisions and business terms, with the detected identifiers replaced. ChatGPT can still analyze the document; it just cannot see the identifiers that were removed.

Step 4: Minimize the Data You Share

Even after PII redaction, practice data minimization — a principle recommended by both the NIST Privacy Framework and the CCPA. Only share the specific content ChatGPT needs for the task:

• Instead of: Uploading an entire 30-page contract for one question about the termination clause
• Do this: Copy just the termination clause and paste it (after redacting any PII in that section)

Less data shared means less data exposed. If ChatGPT only needs three paragraphs, do not give it thirty pages.

Step 5: Do Not Include Sensitive Context in Prompts

Many users inadvertently expose data through their prompts, not just their documents. Avoid writing prompts like:

• "Review this contract between John Smith and Acme Corp" — use "Review this contract between [PARTY A] and [PARTY B]"
• "My client at 123 Main Street, Springfield..." — use "My client at [ADDRESS]..."
• "The employee earning $150,000 at..." — use "The employee earning [SALARY] at..."

Your prompts are part of the data ChatGPT collects and may retain. Apply the same redaction discipline to your instructions as you do to your documents.

ChatGPT vs. Other AI Tools: Privacy Comparison

ChatGPT is the most popular AI tool, but it is not the only one. Understanding how competing platforms handle data helps you choose the safest option for sensitive work — or more importantly, understand that PII redaction reduces your exposure whichever platform you use.

Justee's analysis starts from the same premise for every platform: to protect data before ChatGPT or any other AI tool, the identifiers should be gone before the upload, not after.

Justee's PII redaction tells you how many items it found and removed, by type, so you can check the redacted copy before the document goes to any of the platforms below.

According to Justee, the habit that does the most to protect data before ChatGPT is the same one that works for Claude, Gemini or Copilot: redact first, then share only what the task needs.

What the Regulations Say About Sharing Data With AI

Beyond ChatGPT's own policies, multiple regulatory frameworks govern what happens when you share personal data with AI platforms. Ignorance of these requirements is not a defense.

FTC Section 5

Under Section 5 of the FTC Act, the Federal Trade Commission can act against unfair or deceptive practices, and it has brought many cases against companies whose data security it considered unreasonable. The FTC has also said there is no AI exemption to the laws on the books, so sharing consumer PII with an AI platform without adequate safeguards can raise the same concerns as sharing it with any other service provider.

CCPA/CPRA

California's Consumer Privacy Act requires businesses to minimize data collection and protect personal information shared with service providers. If your business is covered by the CCPA and you upload consumer data to ChatGPT, its data minimization principle applies. Sharing more PII than the task needs can conflict with that principle. Redacting PII before upload is one practical way to share less.

NIST AI Risk Management Framework

The NIST AI RMF identifies privacy as a core AI risk category and recommends data minimization as a foundational control. NIST frameworks are not legally binding for most private organizations, but they are widely used as a reference point when organizations decide what "reasonable" data protection looks like.

Industry-Specific Requirements

If you work in a regulated industry, additional rules apply:

• Healthcare: Sharing patient information with AI tools may implicate HIPAA's minimum necessary standard
• Financial services: The Gramm-Leach-Bliley Act requires protection of customer financial data shared with third parties
• Education: FERPA restricts disclosure of student education records to third-party services
• Legal: ABA Model Rule 1.6 requires lawyers to make reasonable efforts to protect client confidentiality when using technology tools

The common thread: these frameworks expect you to protect personal data when you share it with third-party services, including AI platforms. PII redaction is one of the simplest, most direct safeguards you can add. For a detailed guide on lawyer-specific obligations, see our PII redaction guide for lawyers.

What to Do If You Already Shared Sensitive Data With ChatGPT

If you have already uploaded documents containing PII to ChatGPT, take these steps immediately:

Step 1: Delete the Conversation

Open ChatGPT, find the conversation containing sensitive data, and delete it. Deleting it from your interface does not erase it from OpenAI's servers right away; under OpenAI's stated policy at the time of writing, deleted chats may be kept for up to 30 days.

Step 2: Verify Your Data Controls

Go to Settings > Data Controls and confirm that "Improve the model for everyone" is toggled off. If it was on when you shared the data, that data may have already entered training pipelines.

Step 3: Submit a Data Deletion Request

Under the CCPA, GDPR, and other privacy laws, you may have the right to request deletion of your personal data. Contact OpenAI's privacy team through their privacy policy contact information and request deletion of the specific data. Be specific about what was shared and when.

Step 4: Assess Notification Obligations

If the data belongs to clients, patients, employees, or consumers — not yourself — assess whether the disclosure triggers notification obligations under applicable privacy laws. Some state breach notification laws set fixed deadlines once a breach is discovered, and whether an upload counts as a breach depends on the law and the facts. Talk to a lawyer if you think notification may be required.

Step 5: Implement Redaction Going Forward

The most important step: implement a PII redaction workflow for all future AI interactions. Use Justee's free redaction tool or another automated solution to strip PII before upload, then review the result. Make it a habit: a few minutes of redaction up front is far easier than the cleanup described above.

Justee PII Redaction is built for this step: use it to protect data before ChatGPT sees a document, then check the redacted copy against your original before you upload.

Max Zaykov is the founder of Justee.ai, an AI tool that helps people understand their legal documents.

The information provided is for educational purposes only and does not constitute legal advice. ChatGPT's data policies and features are subject to change by OpenAI — verify current practices directly at openai.com before making data handling decisions. Privacy law requirements vary by jurisdiction and industry. Consult a qualified privacy attorney for advice specific to your situation.

"Justee's analysis replaces names, common SSN formats, addresses and account numbers with numbered placeholders, which is the simplest way to protect data before ChatGPT reads a document."

"PII detection in Justee's redaction tool is best-effort and may not catch every identifier, so checking the redacted copy against your original is part of how you protect data before ChatGPT, not an optional extra."

Related resources: AI contract review, document comparison tool, AI contract review guide, free AI contract review tools.