On this page10 sections
Key takeaways
- ChatGPT Free, Plus, and Pro accounts have model-training data sharing enabled by default (OpenAI Help Center, Nov 2026).
- Claude users who opt in to training have data retained up to 5 years in de-identified form; opt-out users get 30-day deletion (Anthropic, effective Sept 28, 2025).
- Gemini chats reviewed by human reviewers are kept up to 3 years even after you delete your activity (Google Gemini Apps Privacy Hub).
- Teal states it does not use resumes, cover letters, or job descriptions to train AI models (privacy policy updated Nov 11, 2026).
- Deleted ChatGPT data is removed within 30 days, but de-identified copies used for training can persist afterward (OpenAI Privacy Policy).
Uploading a resume to an AI tool sends your work history and contact details to another company’s servers. Depending on the tool, plan, and settings, the conversation may be stored, reviewed by people, or used for model training. Check those settings and remove sensitive details before uploading.
Storage, retention, human review, and training are separate questions. This guide compares the policies dated November 14, 2026, and shows where to change the relevant settings.
What Actually Happens to Your File, Step by Step
When you upload a resume to an AI chatbot, the file leaves your device, gets converted to text on the company’s servers, and becomes part of a stored conversation attached to your account. Everything that happens after that is decided by your plan and your settings.
The tool processes the file’s contents on the company’s servers. Depending on the product and settings, the upload and conversation may remain in your account history.
The conversation may also be retained after deletion, sampled for human review, or used for model training. Check each policy separately; turning off one use does not necessarily stop the others.
Compare Training and Retention Settings
The short version: ChatGPT’s personal plans use your chats for model training by default, Claude makes you choose, and Gemini can use your chats to improve its services whenever activity saving is on. The business tiers of ChatGPT and Claude are excluded from training by default.
| Platform | Training policy | Retention | Where you control it |
|---|---|---|---|
| ChatGPT (Free, Plus, Pro) | Yes, “Improve the model for everyone” is on by default | Deleted data removed from OpenAI’s systems within 30 days; temporary chats kept up to 30 days | Settings → Data Controls |
| ChatGPT (Business, Enterprise, Edu) | No, excluded by default | Ask your workspace administrator which retention policy applies | Workspace admin settings |
| Claude (Free, Pro, Max) | Your choice since Sept 28, 2025; you opt in or opt out | Opt in: up to 5 years, de-identified; opt out: deleted within 30 days. Safety exceptions are explained below. | Settings → Privacy |
| Gemini Apps | When Gemini Apps Activity is on, chats may be used to improve Google services | Auto-delete default is 18 months; with Activity off, chats are kept 72 hours. Reviewed chats can remain for up to 3 years. | Gemini Apps Activity setting |
| Resume builders (Teal, Enhancv, Rezi) | Check the builder’s policy; Teal says it does not train on career content | Check the builder’s deletion policy; Enhancv says it keeps data only “as long as necessary” | Per-tool account settings |
ChatGPT: the default is on, and paying doesn’t change it
OpenAI’s help center says: “On personal workspaces with ChatGPT Free, Plus, or Pro, data sharing is enabled by default.” That’s from OpenAI’s data controls page. Upgrading to Plus or Pro doesn’t flip the default; only Business, Enterprise, and Edu workspaces are excluded from training automatically.
Two details in OpenAI’s policies matter. First, if you delete anything, OpenAI’s privacy policy says it’s removed from their systems within 30 days, but content that was “de-identified and disassociated from your account” for model improvement can persist. Second, the temporary chat feature stays out of your history, isn’t used for training while it’s temporary, and is kept up to 30 days for safety.
Submitting feedback can change how a conversation is used. Even after you turn training off, OpenAI’s help center says that if you submit feedback (a thumbs up or thumbs down) “the entire conversation associated with that feedback may be used to train our models.” Avoid submitting feedback on conversations that contain information you don’t want used for training.
Claude: you choose, but the pop-up nudges you
Anthropic changed its consumer terms in an announcement dated August 28, 2025, effective September 28, 2025. Users on Free, Pro, and Max plans now decide whether their chats and coding sessions can be used for model training. If you opt in, Anthropic retains that data (de-identified) for up to 5 years, per its announcement and privacy policy update. If you opt out, the old 30-day deletion window applies.
Anthropic’s privacy center adds fine print: deleting a chat removes it from history immediately and from back-end storage within 30 days, and it won’t train future models. But chats flagged for policy violations are kept up to 2 years, and trust-and-safety classification scores up to 7 years. Business products (Claude for Work, Government, Education, and the API) are excluded from all of this.
TechCrunch’s coverage from August 28, 2025 noted: the acceptance pop-up’s training toggle was “automatically set to ‘On,’” with the opt-out easy to miss. So if you clicked through that dialog on a busy afternoon, check Settings → Privacy and see what you actually picked.
Gemini: human review, and a 3-year memory for it
Google’s Gemini Apps Privacy Hub lays out the most complicated retention picture of the three. Activity auto-deletes after 18 months by default; you can change that to 3 months, 36 months, or never. Turn Gemini Apps Activity off, and chats are kept only 72 hours, and not used to train AI models.
But Google also samples a subset of conversations for human review. Chats that get reviewed are disconnected from your account, and that’s where the retention gets long:
Even after you delete your Gemini Apps activity, conversations that were reviewed by human reviewers, and related data like feedback are retained by Google for up to three years, because they’re no longer tied to your account.
Deleting your activity does not delete copies that Google has separated from your account for human review.
The Four Pieces of Your Resume That Actually Matter
Your resume is a bundle of personally identifiable information, and it’s the combination (not any single field) that makes it sensitive. PII (personally identifiable information) is any data that can identify you on its own or combined with other details: your name, phone number, email, home address, and employment history.
When you upload the full file, all of it goes at once:
- Phone number and email. These are what scrapers and spam lists want. A chat record containing both, sitting on a third-party server for months, is one more copy of them in the world.
- Street address. Older resume formats still include full mailing addresses. City and state do the same job for location; the street line adds nothing.
- Your employment timeline. Employer names, job titles, and dates together are often enough to identify you even with the name removed. This matters if a tool claims to “anonymize” your data; de-identifying a career history is harder than it sounds.
- Things that shouldn’t be on a resume at all. Date of birth, Social Security number, or a photo. If any of those are in your file, strip them before the file goes anywhere.
Remove details the tool does not need for the task. You can usually get feedback on wording without including your phone number, full address, or client names.
Keep Your Employer’s Confidential Information Out of Prompts
Check what you paste alongside your resume. Internal decks, unreleased product names, client lists, and confidential financial figures may belong to your employer. Leave those details out of a personal AI account.
Remember whose data that is. Your employer likely has legal obligations around that information, and you don’t get to make the disclosure decision for them. If you’re offloading parts of your search to automation, this compounds; AI job search agents touch far more than a single resume.
What About Dedicated Resume Builders and Job Boards?
The niche tools vary more than the big three, and a couple make unusually clear commitments.
Teal’s privacy policy, updated November 11, 2026, states: “Teal does not use your career content, such as resumes, cover letters, and job descriptions, to train AI models.” It does send resume content to third-party AI processors (for document parsing and job matching) but says those providers process only to return results and don’t train on it. For one Gmail-based feature, content routes through OpenRouter to Claude on Amazon Bedrock on a zero-data-retention path.
Enhancv’s privacy policy, last updated July 23, 2024, lists OpenAI among its “AI Technologies” vendors and says vendors “cannot do anything with your personal information unless we have instructed them to do it.” It also states the company hasn’t sold or shared personal information in the preceding 12 months, and keeps data only as long as necessary. Note the policy is older than the big platforms’ current versions, so read it fresh before relying on it.
Rezi’s privacy policy requires only an email address to sign up and says that if you build a resume without an account, “no content from that resume is ever stored on any servers.” It also says it doesn’t sell users’ private personal information, and closing your account from the dashboard permanently deletes your data.
Then there’s the place your resume actually lives: LinkedIn. On November 3, 2025, LinkedIn’s updated terms extended the use of member data (profile details and public content, not private messages) to train content-generating AI models in more regions. The “Data for Generative AI Improvement” setting controls it, opting out doesn’t reduce your access to AI features, and UK, Canada, and Hong Kong members’ setting also governs sharing with Microsoft. Check it. Whether AI-written applications themselves cross an ethical line is another question; that question has its own guide.
Eight Ways to Protect Yourself Before You Upload Anything
Before uploading, review your settings and remove details the tool doesn’t need:
- Turn off training on ChatGPT. Settings → Data Controls → toggle off “Improve the model for everyone.” Your history stays; the training use stops.
- Make your Claude choice deliberately. Settings → Privacy. Don’t just click “Accept” on the terms pop-up; that’s how the toggle ends up On.
- Turn off Gemini Apps Activity before sensitive chats. With Activity off, chats are kept 72 hours and not used to train models. Or use temporary chats in any of the three tools.
- Redact before you upload. Replace your phone number, street address, and email with placeholders like [PHONE]. The AI doesn’t need them for feedback on content or formatting.
- Strip anything your employer would call confidential. Client names, unreleased products, internal metrics. Rewrite them as “a Fortune 500 retailer” or “a Series B fintech.”
- Use temporary chat modes for one-off reviews. ChatGPT’s temporary chats aren’t used for training and vanish from your history; Gemini’s equivalent keeps data only 72 hours.
- Skip the thumbs-up on sensitive conversations. Feedback can pull the whole conversation back into training even after you’ve opted out.
- Delete the conversation when you’re done (full clearance takes up to 30 days) and check the AI toggles on any job board you use.
For how these three handle resume editing tasks beyond privacy, see ChatGPT vs. Claude vs. Gemini for resume editing.
Your Rights Under CCPA and GDPR, The Short Version
If you’re a California resident, the CCPA gives you real leverage. Under the California Attorney General’s CCPA page, you can request to know what personal information a business holds (twice a year, free), request deletion, correct inaccuracies, and opt out of sale or sharing, and businesses must respond within 45 calendar days.
Outside the US, the rules lean harder on the companies and employers. The UK’s ICO recruitment guidance requires employers to handle candidate data (CVs included) under UK GDPR and the Data Protection Act 2018, with a dedicated section on automated decision-making in hiring. Practically: in the US you can send deletion requests to the AI company itself, and in Europe your employer also owes you careful handling of what you sent them.
Retention Matters More Than Training
Training gets all the headlines, but retention is where your data actually lives. A model trained on de-identified resume text is an abstract harm; a chat record with your phone number sitting on a server for years is a concrete one. And the de-identification every policy leans on is weaker than it sounds for career histories, which are nearly self-identifying.
The other thing that surprised me most: the feedback loophole. You can opt out of training on all three platforms and then undo it with one careless thumbs-down. If the tools annoy you mid-review, close the chat instead of rating it.
And let me push back on the opposite take too; “just don’t use AI” isn’t a plan in 2026, any more than “just don’t use email” was in 2005. The checklist above takes five minutes and covers the real exposure. Use the tools; stop donating data you didn’t mean to give.
FAQ
Do AI companies train their models on my resume?
On ChatGPT’s Free, Plus, and Pro plans, yes, by default; you have to turn “Improve the model for everyone” off. Claude asks you to opt in or out, Gemini can use chats to improve its services when activity saving is on, and Teal explicitly says it doesn’t train on resume content.
Does deleting a chat remove my resume from the AI’s servers?
Mostly, but not instantly and not always. OpenAI removes deleted data within 30 days and Anthropic deletes chats from back-end storage within 30 days, while Google keeps chats that were reviewed by human reviewers for up to 3 years even after you delete your activity.
Do humans actually read my chats?
Yes, sometimes. Google says a subset of Gemini chats goes to trained human reviewers to improve quality and safety; Anthropic and OpenAI reserve the right to review content flagged for policy violations.
Is it safer to pay for a subscription?
For team tiers, yes; ChatGPT Business, Enterprise, and Edu are excluded from training by default. But ChatGPT Plus and Pro personal accounts still have training data sharing on by default, so paying alone doesn’t protect you.
What should I remove from my resume before uploading it to AI?
Remove your phone number, street address, email, date of birth, and Social Security number, replacing them with placeholders. The AI can critique your content and formatting without any of them.
Do AI resume builders train on my resume?
Teal explicitly states it does not use resumes, cover letters, or job descriptions to train AI models. Enhancv sends resume content to vendors like OpenAI under instructions and says it hasn’t sold personal information in the preceding 12 months; Rezi says it doesn’t sell users’ private personal information.
Can I get my resume deleted under GDPR or CCPA?
Yes. California residents can request deletion under the CCPA and businesses must respond within 45 days, while EU and UK residents have equivalent access and erasure rights under GDPR, submit requests through each company’s privacy portal.
Is LinkedIn training AI on my profile and resume?
Yes, since November 3, 2025, LinkedIn uses profile details and public content to train content-generating AI models in more regions, with the setting on by default. Turn it off under “Data for Generative AI Improvement” in Settings; opting out doesn’t limit your access to LinkedIn’s AI features.
Sources
- Data Controls in ChatGPT: OpenAI Help Center
- How your data is used to improve model performance: OpenAI Help Center
- Temporary chat in ChatGPT: OpenAI Help Center
- Privacy Policy: OpenAI
- Updates to our Consumer Terms: Anthropic
- Updates to our Privacy Policy: Anthropic Privacy Center
- How long do you store my data?: Anthropic Privacy Center
- Anthropic users face a new choice: opt out or share your data for AI training: TechCrunch
- Gemini Apps Privacy Hub: Google Support
- Privacy Policy: Teal
- Privacy Policy: Enhancv
- Privacy Policy: Rezi
- Update to our Terms and data use: LinkedIn Help
- California Consumer Privacy Act (CCPA): California Department of Justice
- Recruitment and selection: UK ICO employment practices guidance
Written by ResumeFlex. Questions or corrections? Contact ResumeFlex. Read our source and correction standards.
Filed under
AI Tools