Can My Employer See My ChatGPT Prompts? A 2026 Guide

Short answer: It depends on three things — which device you are on, which network you are on, and which ChatGPT account you are signed into. On a personal device, on a personal network, signed into a personal ChatGPT account, your employer almost certainly cannot see your prompts. On a work device, work network, or a ChatGPT Enterprise / Business account your employer manages, the answer ranges from "they can if they want to" to "they already have a record of every prompt you sent." This guide explains, in plain English, how each pathway actually works in 2026, what is and isn't visible, and what you can do about it.

The three pathways your employer might use

1. The account itself

If you are signed into ChatGPT Enterprise, ChatGPT Business (formerly Team), ChatGPT Edu, or a company-managed Microsoft 365 Copilot, Google Workspace Gemini, or Anthropic Claude for Work account, your employer is the account administrator. Admins on those plans can — depending on the product and how it's configured — see:

• Which users have used the tool, when, and how often.
• The titles of conversations.
• The full content of conversations, in many cases, via admin export, compliance APIs, Microsoft Purview, or Google Vault. (On Claude, content access is through the Compliance API on the Enterprise plan; Anthropic's audit logs do not include chat content.)
• File uploads, custom GPTs, projects, and connectors used.

This is by design — these are enterprise products built for organizations that need to supervise AI use. The terms of service make it clear, but most users do not read them.

If you are signed into a personal ChatGPT Free / Plus / Pro account, your employer is not the account administrator and has no direct visibility into the conversations through OpenAI. They can still see the activity through pathways 2 and 3 below.

2. The device

If you are using a work-issued laptop, phone, or VDI (virtual desktop), the device itself is typically managed by your employer. Modern endpoint management can include:

• Endpoint DLP (data loss prevention) that inspects what you type or paste into AI chat boxes.
• Endpoint detection and response (EDR) software that records process activity and, in some configurations, screen content.
• Browser extensions deployed via group policy that intercept AI traffic, log prompts, or redact sensitive data.
• Screen recording or session monitoring software (rare in regular office work, common in regulated environments like financial trading floors and call centers).
• Configuration profiles (Mobile Device Management, Microsoft Intune, Jamf) that can install monitoring agents, restrict apps, and pin browsers to a managed identity.

If the device is yours, none of these apply unless you knowingly install them.

3. The network

If you are connected to your employer's corporate network, VPN, or zero-trust gateway, traffic to AI services may pass through a system that can:

• Log the destination domain (chatgpt.com, claude.ai, gemini.google.com) and the time of access.
• Inspect the contents of HTTPS traffic via TLS interception (the corporate root certificate trusted by your work device is what makes this possible).
• Block, allow, or coach AI traffic via a CASB (Cloud Access Security Broker) like Netskope, Zscaler, Palo Alto, Microsoft Defender for Cloud Apps, or similar.

If you are on a personal network with no work-mandated VPN active, the corporate network has no visibility. If a corporate VPN is connected, the corporate network has visibility as if you were in the office, even from home.

What this combines to in practice

The most-misunderstood case is the bottom row: "incognito" mode does not hide activity from a managed device. Endpoint agents see process activity at the operating system level, below the browser. Network-layer controls see DNS lookups and TLS handshakes regardless of whether the browser tab is incognito.

What OpenAI itself stores (separate from your employer)

Even on a personal account on a personal device, your prompts do not vanish into thin air. As of our November 2026 review:

• ChatGPT Free / Plus / Pro: Conversations are retained according to your Data Controls settings. Even in a Temporary Chat (the successor to the old "chat history off" switch), OpenAI keeps the conversation for up to 30 days and may review it for abuse before deletion. On Free, Plus and Pro, prompts can be used to train future models unless you turn off "Improve the model for everyone" in Data Controls.
• ChatGPT Business / Enterprise / Edu: Prompts are not used to train models by default. Workspace admins control how long conversations are retained; deleted conversations are removed from OpenAI's systems within 30 days unless OpenAI is legally required to keep them.
• Temporary Chat: Per OpenAI's product documentation, Temporary Chats are not used for training and are retained only as long as needed for abuse monitoring (currently up to 30 days). They are not "off-the-record" — they are "do-not-store-in-history."
• OpenAI API: Default 30-day retention for abuse monitoring; Zero Data Retention available for eligible enterprise customers.

A government subpoena, civil discovery order, or qualifying court order can require OpenAI to produce stored prompts. This is true on every tier — including consumer. In 2025 a court order in the New York Times case required OpenAI to preserve consumer ChatGPT and non-ZDR API content — including deleted and temporary chats — for several months; Enterprise and Edu were excluded. OpenAI says the obligation ended on September 26, 2025.

A company you are applying to is a different matter entirely, with a much shorter answer: see Can an Interviewer See Your ChatGPT or Claude History?.

What about ChatGPT, Claude, and Gemini all at once?

The same three pathways apply to every browser-based AI assistant:

• Claude — Claude Enterprise has audit logs (metadata only) and a Compliance API that can pull chat and file content; Team plans have admin controls but not these features. Claude.ai consumer is invisible to your employer's admin console but still subject to the device and network pathways. The account-by-account version of this question is Can My Employer See My Claude Chats?.
• Gemini — Gemini app conversations in a Workspace account can be searched, exported, retained and placed on hold by admins in Google Vault (search and export since February 2025; retention rules and holds since June 2026); Gemini features inside Docs or Gmail are covered by audit logs rather than Vault search. Consumer Gemini is invisible to admins but, again, subject to device and network monitoring on managed devices. Can My Employer See My Gemini Chats? walks through the Workspace side in detail.
• Microsoft Copilot — M365 Copilot prompts and responses are captured in the Microsoft Purview unified audit log, with rich event capture for tenant admins. Personal Copilot at copilot.microsoft.com is not visible to admins but is subject to device and network controls on managed Windows devices; Purview can detect it as a third-party AI app through browser activity on onboarded devices.

The browser-extension picture is similar: an extension installed via group policy on a managed browser sees and acts on every prompt regardless of which AI service it goes to.

How to tell what is monitored on your device

You don't always need to ask. A few quick checks on a Windows or macOS work device:

• Check installed browser extensions. If your employer pushed an "AI policy" extension or a CASB extension, it is listed in your browser's extension manager and may not be removable.
• Check the certificate on chatgpt.com. Click the lock icon → Connection is secure → Certificate. If the issuer chain ends at a name like "Acme Corp Root CA" rather than DigiCert, Sectigo, or Google Trust Services, your network is performing TLS interception.
• Check installed apps and configuration profiles. On macOS, System Settings → Privacy & Security → Profiles. On Windows, Settings → Accounts → Access work or school. The presence of an MDM profile means the device is managed.
• Check your IT acceptable-use policy. Most employers explicitly state in the AUP that work devices and networks are monitored. In the US, federal wiretap law allows interception where one party to the communication has consented, and several states — New York and Delaware among them — require employers to give notice of electronic monitoring. The protection for the employer comes from the disclosure, not from secrecy.

If you genuinely do not want your employer to see what you put into AI — even for legitimate personal reasons — use a personal device on a personal network, and sign in to a personal account.

What you can do about it

The honest answer is that most employees should treat their work AI use the way they treat their work email: assume it is logged, and behave accordingly. For specific concerns:

• For workforce AI use that should be private (counseling, job-searching, personal medical questions): use a personal device on a personal network. Don't blur the boundary.
• For workforce AI use that involves sensitive customer or company data: use the company-sanctioned tool, not a personal account on a work device. Personal accounts on a managed device offer the worst of both worlds — your employer often can see, and you've also routed regulated data through a non-BAA, non-DPA-covered account.
• For employers who don't want to inspect prompt contents but do need to keep regulated data off the wire: use a local-first privacy layer that detects sensitive entities and replaces them with one-way placeholders on the device, before any prompt leaves. Locke runs on the user's device as a desktop app, with a companion browser extension for AI websites and a local proxy for command-line, script and SDK traffic. Locke 1.0 replaces detected values with one-way placeholders before the prompt leaves the device, so the original values are not sent to the AI provider or to Sonomos. Coverage depends on the app, platform and settings, and it does not stop a managed device from logging what the user types.

Frequently asked questions

Can my employer see my ChatGPT prompts if I'm on a personal device but using their VPN?

The corporate VPN gives them visibility into network destinations and timing. Whether they can see the content depends on whether they perform TLS interception (uncommon for personal-device VPN configurations) and whether they have a browser extension or endpoint agent on the personal device (which would require you to install it). For most BYOD VPN setups, your employer can see that you went to chatgpt.com and when, but not what you typed.

Does using ChatGPT in incognito hide it from my employer?

No, not on a managed device. Incognito mode prevents the browser from storing local history and cookies, but it does not hide traffic from network monitoring, endpoint DLP, EDR, or push-installed extensions. On a personal device on a personal network, incognito hides nothing more than your local history — your employer wasn't seeing it either way.

Can my employer read my chats in ChatGPT Enterprise?

Yes, depending on configuration. Workspace owners and admins can access conversation logs and content through the admin console and the Compliance API (the "Conversation messages" permission can be granted only by a workspace owner), particularly when responding to legal hold, e-discovery, or internal investigation requests. The exact rights and processes depend on the contract and your employer's policy. Treat ChatGPT Enterprise like work email.

Can my employer see what I wrote in ChatGPT Team?

ChatGPT Business (formerly Team) conversations are not used for training by default, and OpenAI states that workspace admins can view, access, export, and delete members' conversations. The Compliance API for SIEM and e-discovery integrations is an Enterprise and Edu feature. Again, treat it like work email.

Can my employer subpoena my personal ChatGPT account?

In litigation involving you and your employer, your personal ChatGPT history can be subject to discovery if it is reasonably calculated to lead to admissible evidence. Whether OpenAI produces it depends on the legal process; whether your employer obtains it depends on the discovery rules in your jurisdiction. Don't assume "personal" means "outside discovery."

What about ChatGPT Memory? Can my employer see what it stored?

Memory is per-account. On a personal account, only you (and OpenAI) can see what's stored. On a Business or Enterprise workspace, admins control whether Memory is available, and anything Memory derives from workspace conversations is subject to the same admin and compliance access as those conversations.

How does Locke relate to all this?

Locke is a privacy layer that runs on your device as a desktop app; its companion browser extension screens AI websites, and a local proxy screens command-line, script and SDK traffic. Locke 1.0 detects sensitive data — names, account numbers, health terms, source-code secrets — and replaces it with one-way placeholders before the prompt leaves the device; reversible tokenization is planned shortly after 1.0. Because the masking happens locally, the detected values are not sent to OpenAI or to Sonomos. Coverage depends on the app, platform and settings. Locke is not a way to hide AI use from your employer; it is a way to keep regulated data out of the prompt in the first place. If your concern is "I want to use AI without leaking client data," Locke directly addresses it. If your concern is "I don't want my employer to know I use ChatGPT at all," the answer is to use a personal device on a personal network, not a tool.

The bottom line

In 2026, the question "Can my employer see my ChatGPT prompts?" has three answers stacked on top of each other: through the account, through the device, and through the network. On a personal device, personal network, and personal account, the answer is essentially no. As soon as any one of those three becomes employer-controlled, monitoring becomes possible — and on a fully managed setup, monitoring is usually the default. Treat your work AI use like your work email: assume it's logged, keep regulated data out of the prompt, and don't paste anything into a personal account that you wouldn't want subpoenaed.

Sources and primary references

Reviewed November 2, 2026. Regulations are cited to their official text. Vendor statements are cited to the vendor's own documentation as it read on that date; vendors revise these pages, so confirm before relying on them.

• Enterprise privacy at OpenAI — what Enterprise and Business admins can access, admin-controlled retention, 30-day deletion, no training by default.
• OpenAI Compliance Platform for Enterprise and Edu customers — Compliance API scope (Enterprise and Edu), owner-granted "Conversation messages" permission, SIEM and CASB partners.
• ChatGPT Team is now ChatGPT Business — rename on August 29, 2025.
• OpenAI Data Controls FAQ — "Improve the model for everyone" toggle; Temporary Chats deleted after 30 days and reviewable for abuse.
• OpenAI Memory FAQ — workspace owners and admins manage memory settings; temporary chats do not create memories.
• OpenAI: How we're responding to The New York Times' data demands — 2025 preservation order scope (consumer and non-ZDR API; not Enterprise or Edu) and its end on September 26, 2025.
• OpenAI API: Your data — 30-day abuse-monitoring logs and zero-data-retention eligibility.
• Microsoft Purview data security and compliance protections for generative AI apps — Copilot prompts and responses in the unified audit log; endpoint DLP blocking pastes into ChatGPT; consumer Copilot detected through browser activity.
• Data, privacy, and security for Microsoft Copilot — Copilot interaction history is stored and viewable by admins via Purview.
• Google Vault now supports Gemini — Vault search and export for Gemini app data from February 7, 2025.
• Use Vault to search Gemini app — Vault returns Gemini app prompts and responses for export.
• Google Vault now supports retention rules and litigation holds for Gemini app — admin retention and holds for Gemini app data (June 2026).
• Anthropic: Access the Compliance API — Enterprise-only API exposing chat data and file content; enabled by the Primary Owner.
• Anthropic: Access audit logs — Enterprise-only audit logs that exclude chat content.
• 18 U.S.C. § 2511 — federal wiretap prohibition and the one-party consent exception.
• New York Civil Rights Law § 52-c — written notice and acknowledgment of electronic monitoring; posting; penalties.
• 19 Del. C. § 705 — Delaware notice requirement for monitoring email and Internet usage; $100 per violation.