Once we work together with synthetic intelligence, we frequently share a big quantity of non-public data with out giving it a lot thought. This data can vary from dietary preferences and marital standing to our dwelling deal with and even social safety quantity. To make sure the safety and privateness of this extremely delicate data, it’s important to know precisely what the AI does together with your information: the place it shops it and whether or not it makes use of it for coaching.
On this submit, we take a more in-depth have a look at the information assortment coverage of one of the crucial in style AI apps, ChatGPT, and clarify methods to configure it to maximise your privateness and safety to the extent that OpenAI permits it. It is a lengthy information — however a complete one.
Desk of contents
What information ChatGPT collects about you
OpenAI, the proprietor and developer of ChatGPT, maintains two privateness insurance policies. The particular coverage that applies to customers is dependent upon the area the place that particular person registered their account:
As a result of these insurance policies are related, we’ll first cowl the frequent parts, after which focus on the variations.
By default, OpenAI collects an in depth array of non-public data and technical information about units from all ChatGPT customers.
- Account data: title, login credentials, date of beginning, billing data, and transaction historical past
- Person content material: prompts in addition to uploaded information, photographs, and audio
- Communication data: contact particulars the person offered when reaching out to OpenAI through electronic mail or social media
- Log information: IP deal with, browser kind and settings, request date and time, and particulars about how the person interacts with OpenAI companies
- Utilization information: details about the person’s interplay with OpenAI companies, comparable to content material considered, options used, actions taken, and technical particulars like nation, time zone, system, and connection kind
- Gadget data: system title, working system, system identifiers, and the browser used
- Location data: the area decided by the IP deal with, relatively than the precise location
- Cookies and related applied sciences: crucial for service operation, person authentication, enabling particular options, and guaranteeing safety; the whole listing of cookies and their respective retention durations is offered right here
What precisely OpenAI does with the information it collects from particular person customers can be mentioned within the subsequent a part of this submit. Right here, we point out the important thing distinction between the privateness insurance policies for customers from the European Financial Space (EEA) and people from different areas. European customers have the best to object to using their private information for direct advertising and marketing. They could additionally problem information processing the place the corporate justifies this by its “respectable pursuits”, comparable to inner administration or enhancements to companies.
Word that OpenAI’s dealing with of information for enterprise accounts is ruled by separate guidelines that apply to ChatGPT Enterprise and ChatGPT Enterprise subscriptions, in addition to API entry.
What OpenAI does together with your information, and whether or not ChatGPT is educated in your chats
By default, ChatGPT can prepare its fashions on person prompts and the content material that customers add. This coverage applies to customers of each the free model and the Plus and Professional subscriptions.
For enterprise accounts — particularly ChatGPT Enterprise, ChatGPT Enterprise, and API entry — coaching on person information is disabled by default. Nevertheless, within the case of the API (the applying programming interface that connects OpenAI fashions to different functions and companies — the only use case being ChatGPT-based buyer help bots), the corporate offers builders with the choice to voluntarily allow information transmission.
OpenAI outlines a complete listing of main functions for processing customers’ private data:
- To maintain companies: to reply to queries and help customers
- To improve and develop companies: so as to add new options and conduct analysis
- To communicate with customers: to inform customers about adjustments and occasions
- To protect the safety of companies: to stop fraud and guarantee safety
- To comply with authorized obligations: to guard the rights of customers, OpenAI, and third events
The corporate additionally states that it could anonymize customers’ private information, although it doesn’t obligate itself to take action. Moreover, OpenAI reserves the best to switch person information to 3rd events — particularly its contractors, companions, or authorities businesses — if such switch is critical for service operation, compliance with the legislation, or the safety of rights and safety.
As the corporate notes on its web site: “In some instances, fashions could be taught from private data to know how parts like names and addresses operate in language, or to acknowledge public figures and well-known entities”.
It’s necessary to notice that each one person information is processed and saved on OpenAI servers in the US. Though the extent of non-public data safety could range from nation to nation, the corporate asserts that it applies uniform safety measures to all customers.
Tips on how to forestall ChatGPT from utilizing your information for AI coaching
To disable the gathering of your information throughout the app, click on your account title within the decrease left nook of the display. Choose Settings, then navigate to Information controls.
Within the Information controls part of the ChatGPT settings, you may disable using your prompts for mannequin coaching
In Information controls, flip off the toggles subsequent to the next objects:
- Enhance the mannequin for everybody: disabling this selection prevents using your prompts and uploads (textual content, information, photographs) for mannequin coaching. Turning this off deactivates the 2 objects beneath it
- Embrace your audio recordings: disabling this selection prevents the voice messages from the dictation characteristic from getting used for mannequin coaching. It’s disabled by default
- Embrace your video recordings: this refers back to the characteristic that permits you to embrace a video stream out of your digital camera throughout a voice chat within the ChatGPT apps for iOS and Android. This video stream can also be used for mannequin coaching. You can even disable this selection by the online software. It’s disabled by default
By turning off these settings, you forestall using new information for mannequin coaching. Nevertheless, it’s necessary to appreciate: in case your prompts or content material have been already used for coaching earlier than you disabled the choice, it’s not possible to take away them from the educated mannequin.
On this similar part, you may delete or archive all chats, and in addition request to Export Information out of your account. This lets you verify what data OpenAI shops about you. A knowledge archive can be despatched to your electronic mail. Please word that making ready an export could take a while.
The Delete account possibility can also be obtainable right here. When your account is deleted, solely your private information is erased; data already used for mannequin coaching stays.
Past the in-app settings, you may handle your information by the OpenAI Privateness Portal. On the portal, you may:
- Request and obtain all of your information saved by OpenAI
- Utterly delete your customized GPTs, in addition to your ChatGPT account and the private information related to it
- Ask OpenAI to not prepare the AI in your information. If OpenAI approves your request, the AI will cease coaching on the information you offered earlier than you disabled the Enhance the mannequin for everybody possibility within the settings
- Typically ChatGPT can also prepare on private information from public sources — you may submit a request to cease this as effectively
- Request the deletion of non-public information from particular conversations or prompts
Customers from the European Financial Space, the UK, and Switzerland have extra rights underneath the GDPR. The legislation is in impact in European international locations and regulates how corporations acquire and use private information. These rights aren’t immediately displayed on the OpenAI Privateness Portal, however they are often exercised by submitting a request by the portal, or by writing to dsar@openai.com.
Tips on how to clear your information from ChatGPT’s reminiscence
One other vital ingredient of privateness safety is ChatGPT’s reminiscence. In contrast to chat historical past, reminiscence permits the mannequin to recall particular particulars about you, comparable to your title, pursuits, preferences, and communication type. This information persists throughout periods and is used to personalize the AI’s responses.
To evaluate precisely what the AI remembers throughout the app, click on your account title within the lower-left nook of the display. Select Settings, then navigate to Personalization, and choose Handle reminiscences.
Below Personalization, you may handle saved reminiscences, briefly disable reminiscence, or forestall the mannequin from referring to speak historical past when responding
This part shows all saved data. If you want for ChatGPT to overlook a selected element, click on the trash can icon subsequent to that reminiscence. Necessary: for a reminiscence to be utterly erased, it is advisable to additionally delete the particular chat the knowledge was saved from. In case you delete solely the chat however not the reminiscence, the information stays saved.
In Personalization, you can even configure what information ChatGPT will retailer about you in future conversations. To do that, you must familiarize your self with the two kinds of reminiscence obtainable within the AI:
- Saved reminiscences are fastened recollections about you, comparable to your title, pursuits, or communication type, which stay within the system till you manually delete them. These are created once you explicitly ask the chat to recollect one thing
- Chat historical past is the mannequin’s capacity to think about particular particulars from previous conversations to provide extra customized responses. On this case, ChatGPT doesn’t retailer each element; as an alternative, it selects solely fragments that it deems helpful. A lot of these reminiscences can change and adapt over time
You’ll be able to disable one or each of those reminiscence varieties within the ChatGPT settings. To deactivate saved reminiscences, flip off the toggle subsequent to Reference saved reminiscences. To do the identical for chat historical past, flip off the toggle subsequent to Reference chat historical past.
Disabling these options doesn’t delete beforehand saved data. The info stays throughout the system, however the mannequin ceases to reference it in new responses. To utterly delete saved reminiscences, go to the Handle reminiscences part as described above.
The Personalization menu within the web-based model of ChatGPT is barely completely different, with a further possibility: Document mode. This permits the AI to reference transcripts of your previous recordings when producing responses. It’s potential to disable this characteristic throughout the internet interface.
As well as, the online model shows a reminiscence utilization indicator, comparable to 87% full, which reveals how a lot area is occupied by reminiscences.
The online model of ChatGPT additionally features a reminiscence utilization indicator underneath Personalization
For delicate conversations, you may make the most of particular Short-term Chats, which the AI received’t keep in mind.
Tips on how to use Short-term Chats in ChatGPT
in ChatGPT are designed to resemble incognito mode in an online browser. If you wish to focus on one thing notably intimate or confidential with the AI, this mode helps scale back the dangers. The chats aren’t saved within the historical past, they don’t change into a part of the reminiscence, and so they’re not used to coach the fashions. This final level holds true for all Short-term Chats whatever the settings chosen within the Information controls part, which was mentioned above.
As soon as a session ends, its contents disappear and can’t be recovered. This implies Short-term Chats received’t seem in your historical past, and ChatGPT received’t keep in mind their content material. Nevertheless, OpenAI warns that for safety functions, a replica of the Short-term Chat could also be saved on the corporate’s servers for as much as 30 days.
In June 2025, a court docket ordered OpenAI to protect all person chats with ChatGPT indefinitely. The choice has already taken impact, and though the corporate plans to attraction it, on the time of this publication, OpenAI is compelled to retailer Short-term Chat information completely in a particular safe repository that “can solely be accessed underneath strict authorized protocols”. This largely nullifies your entire idea of “Short-term Chats”, and confirms the previous adage, “There’s nothing extra everlasting than the momentary”.
It’s necessary to notice that when making a Short-term Chat, you’re beginning a dialog with the AI from a clean slate: the chatbot received’t keep in mind any data from its earlier chats with you.
To provoke a Short-term Chat within the web-based model of ChatGPT, open a brand new chat and click on Activate momentary chat button within the higher proper nook of the web page.
Within the internet model of ChatGPT, the Activate momentary chat button is positioned within the higher proper nook of the display, and launches a brand new chat that received’t save any historical past or reminiscence
To activate a Short-term Chat within the ChatGPT functions for macOS and Home windows, click on the AI mannequin choice, and a Short-term Chat toggle will seem on the backside of the window that opens.
Within the ChatGPT app for macOS, Short-term Chat activation is offered within the mannequin choice menu
After a Short-term Chat is activated, a particular display will open, which can look barely completely different within the desktop and internet variations. In case you see this display, it means issues are working appropriately.
Short-term Chats aren’t saved in historical past, used to replace reminiscence, or utilized for mannequin coaching
Integrating ChatGPT together with your system functions
The ChatGPT software features a characteristic named Work with Apps. This lets you work together with the AI past the ChatGPT interface itself, extending its performance into different apps in your system. Particularly, the mannequin can hook up with textual content editors and numerous improvement environments.
If you make the most of this characteristic, you may obtain AI strategies and make edits immediately inside these apps, eliminating the necessity to copy textual content to a separate chat window. The core idea is to embed the AI into your present, acquainted workflows.
Nevertheless, together with the comfort, this characteristic introduces privateness dangers. By connecting to functions, ChatGPT good points entry to the content material of the information you’re engaged on. These information could embrace private paperwork, work tasks or stories, notes containing confidential data, and different related content material. A portion of this information could also be despatched to OpenAI’s servers for evaluation and response technology.
Subsequently, the extra functions you grant entry to, the upper the chance that delicate data can be uncovered to OpenAI.
On the time of this submit, the ChatGPT software for macOS can hook up with the next functions:
- Textual content-editing and note-taking apps: Apple Notes, Notion, TextEdit, Quip
- Growth environments: Xcode, Script Editor, VS Code (Code, Code Insiders, VSCodium, Cursor, Windsurf)
- JetBrains IDEs: Android Studio, IntelliJ IDEA, PyCharm, WebStorm, PHPStorm, CLion, Rider, RubyMine, AppCode, GoLand, DataGrip
- Command-line interfaces: Terminal, iTerm, Warp, Immediate
No comparable listing has been printed for the Home windows model of the app but.
To verify if this characteristic is at present enabled in your system, click on your account title within the lower-left nook of the display. Choose Settings and scroll all the way down to Work with Apps. If the toggle swap subsequent to Allow Work with Apps is on, the characteristic is turned on.
In Work with Apps, you may verify if the characteristic is enabled, and handle connections to put in apps
It’s necessary to emphasise that enabling the characteristic doesn’t instantly give the ChatGPT app entry to the functions in your system. For ChatGPT to research and make adjustments to content material in different apps, the person should explicitly grant a separate permission to every particular person app.
In case you’re not sure whether or not you’ve granted ChatGPT any entry permissions, you may confirm this throughout the similar part. To do that, choose Handle Apps. The window that opens will show each app in your system that ChatGPT can probably work together with. If every app reveals Requires permissions beneath it, and Allow Permission on the best, it signifies that ChatGPT at present has no entry to any apps.
Handle Apps shows the apps ChatGPT can probably entry
On macOS, do you have to select to grant ChatGPT entry to an software, you could additionally allow the AI app to regulate your laptop through the accessibility options within the system settings. This permission grants ChatGPT in depth additional capabilities: monitoring your actions, managing different functions, simulating keystrokes, and interacting with the person interface. For this very purpose, these permissions are granted solely manually and require the person’s express affirmation.
In case you’re involved concerning the uncontrolled sharing of your information with ChatGPT, we advocate you disable the Allow Work with Apps toggle swap and forgo utilizing this characteristic.
Nevertheless, in order for you ChatGPT to have the ability to work with functions in your system, you must take note of the next three options, and configure them in line with your private steadiness of privateness and comfort:
- Routinely pair with apps from chat bar permits ChatGPT to routinely hook up with supported functions immediately from the chat UI with out requiring handbook choice every time. This hastens your workflow, however will increase the chance that the mannequin will achieve entry to an software that the person didn’t intend to attach it to
- Generate advised edits permits ChatGPT to suggest adjustments to textual content or code throughout the related software, however you’ll want to use these adjustments manually. That is the safer possibility as a result of the person retains management over adjustments being made
- Routinely apply suggested edits permits the mannequin to right away implement adjustments to information. Whereas this maximizes course of automation, it carries extra dangers, as modifications may very well be utilized with out affirmation — probably affecting necessary paperwork or work tasks
Tips on how to join ChatGPT to third-party on-line companies
ChatGPT may also be related to third-party on-line companies for higher customization: this enables the AI to supply extra exact solutions and execute duties higher by contemplating, for instance, your electronic mail correspondence in Gmail or schedule in Google Calendar.
In contrast to Work with Apps, which allows ChatGPT to work together with domestically put in functions, this characteristic includes exterior on-line platforms like GitHub, Gmail, Google Calendar, Groups, and lots of others.
The precise listing of accessible companies is dependent upon your plan. Probably the most in depth choice is offered within the Enterprise, Enterprise, and Edu tiers; a barely extra restricted set is present in Professional; and the roster of companies is considerably extra modest in Plus. Free customers don’t have any entry to this characteristic. Some regional restrictions additionally apply. You’ll be able to view the total listing for all plans by following the hyperlink.
When connecting to third-party companies, it’s essential to know precisely what information OpenAI will course of, how, and for what functions. In case you haven’t disabled coaching in your information, data obtained from related companies can also be used for mannequin coaching. Moreover, with the reminiscence possibility enabled, ChatGPT is able to remembering particulars obtained from third-party companies and using them in future chats.
To view the listing of on-line companies obtainable for connection, click on your account title within the backside left nook of the display. Then, choose Settings and, within the Account part, navigate to Connectors.
Connectors obtainable within the ChatGPT settings
Below Connectors, you’ll see companies which can be already related, in addition to these which can be obtainable for activation. To disconnect ChatGPT from a service, choose the service and click on Disconnect.
The settings for every connector let you disable ChatGPT’s entry to the service, view the date when it was related, and permit or disallow the automated use of its information in chats
To mitigate privateness dangers, we advocate connecting solely the completely crucial companies, and configuring the reminiscence and information controls inside ChatGPT upfront.
Tips on how to arrange safe login to ChatGPT
If you’re a frequent ChatGPT person, the service seemingly shops considerably extra details about you than even social media. Subsequently, in case your account is compromised, attackers may achieve entry to information they will use for doxing, blackmail, fraud, theft of funds, and different kinds of assaults.
To mitigate these dangers, it’s important to set a fancy password, and allow two-factor authentication for logging in to ChatGPT. What we take into account once we say “complicated” is a password that meets all the following standards:
- A minimal size of 16 characters
- A mixture of uppercase and lowercase letters, numbers, and particular characters
- Ideally, no dictionary phrases, no easy sequences like “12345” or “qwerty”, and no repeating characters
- Uniqueness: a special password for every web site or on-line service
In case your present ChatGPT password doesn’t fulfill these standards, we strongly advocate you modify it. Whereas there’s no possibility to vary the password as such within the ChatGPT settings, you should use the password reset process. To do that, sign off of your account, choose Forgot password? on the login display, and comply with the directions to arrange a brand new password.
It’s possible you’ll be tempted to make use of the AI mannequin itself to generate a password. Nevertheless, we don’t advocate this: our analysis means that chatbots are sometimes not very efficient at this activity, and continuously generate extremely insecure passwords. Moreover, even when you explicitly ask the neural community to create a random password, it received’t be actually random, and can subsequently be extra weak.
For extra account safety, we additionally advocate enabling two-factor authentication: navigate to Settings, choose Safety, and activate the Multi-factor authentication toggle swap. After this, scan the QR code in an authenticator software, or manually enter the key key that seems on the display, and confirm the motion with the one-time code.
Within the Safety part of the online model, you can even sign off of all lively periods on all units, together with your present one. Sadly, you can’t view the login historical past. We advocate utilizing this characteristic when you suspect that somebody could have gained unauthorized entry to your account.
Within the internet model’s safety settings, you may allow multi-factor authentication, and in addition sign off of ChatGPT on all units
Last tricks to safe your information
When utilizing AI chatbots, it’s necessary to keep in mind that these functions create new privateness challenges. To guard our information, we now should account for issues that weren’t a priority when organising accounts in conventional apps and internet companies, and even in social media and messaging apps. We hope that this complete information to privateness and safety settings in ChatGPT will enable you to with this difficult activity.
Additionally, please keep in mind to safeguard your ChatGPT account in opposition to hijacking. One of the simplest ways to do that is through the use of an app that generates and securely shops sturdy passwords, whereas additionally managing two-factor authentication codes.
Kaspersky Password Supervisor helps you create distinctive, complicated passwords, autofill them when logging in, and generate one-time codes for two-factor authentication. Passwords, one-time codes, and different information encrypted in Kaspersky Password Supervisor will be synchronized throughout all of your units. It will assist present strong safety in your account in ChatGPT and different on-line companies.
In case you’re in search of extra data on the safe use of synthetic intelligence, listed here are some extra helpful posts:
