Can government employees use ChatGPT? Work tools, personal accounts, and public records

The Morse teamPublished October 31, 202610 min read

Yes, government employees can use ChatGPT, but for government work only through a version their agency has approved, and as of October 31, 2026, most of the federal and state policies we read bar personal accounts for work or block unapproved AI tools on government devices. Federal agencies can buy ChatGPT, Claude, Gemini, and Grok at government prices through GSA, and GSA's own USAi platform offers agencies AI chat at no cost. A personal ChatGPT account is a different matter: it's fine for your own life on your own device, but work done in it can become a public record, and some states say so outright.

We read the governing documents on the government's own sites on October 31, 2026: OMB's AI memos, GSA's OneGov releases and contract listings, GSA's and the Consumer Product Safety Commission's AI policies, the National Archives' August 2026 guidance on AI records, the federal records and ethics rules, Pennsylvania's and Iowa's state AI policies, and Washington State Archives' guidance. Where a primary page wouldn't load, we say so and cite the requester's own reporting. GSA, Pennsylvania, and the National Archives deserve credit for publishing their rules in full, with dates, so an employee can read exactly what applies.

This isn't legal advice. For your own situation, ask your agency's records officer, ethics official, or Chief AI Officer.

Can federal employees use ChatGPT for work?

Yes, if their agency provides it. The governing memo is OMB M-25-21, dated April 3, 2025, which "rescinds and replaces" the 2024 memo M-24-10 and told each agency to "develop a policy that sets the terms for acceptable use of generative AI" within 270 days (OMB M-25-21). That deadline fell at the end of December 2025, so your agency should have its own rule by now. A later memo, M-26-04 of December 11, 2025, governs which large language models agencies may buy and told them to set up a way for "agency users of LLMs to report outputs" that break its neutrality principles by March 11, 2026 (OMB M-26-04).

A guide that still cites M-24-10 as the floor for agency AI use is out of date. M-25-21 took its place on April 3, 2025 (OMB M-25-21).

The tools themselves come mostly through GSA. On August 6, 2025, GSA offered ChatGPT Enterprise to "every participating U.S. federal agency" for "a nominal fee of $1" for one year (GSA). Anthropic followed on August 12 with Claude for all three branches at the same price, a deal GSA said supports "FedRAMP High workloads" for "sensitive unclassified work" (GSA). Two days later GSA launched USAi, a free "secure generative artificial intelligence evaluation suite" for chat, code, and summaries, "available today to all federal agencies" (GSA).

The $1 offers are running out. GSA's OneGov listing, read October 31, 2026, shows:

Source: GSA OneGov. The new ChatGPT terms, announced September 10, 2026, run 27 months from October 1 with "no platform-access fee, no minimum orders, and no spend commitment" (GSA), and Claude's discount followed on October 27 (GSA). Both new offers open to "state, local, and tribal governments," not just federal agencies. Under the new ChatGPT and Claude offers, agencies pay for what they use, so check whether yours renewed before you assume your seat survived the switch.

Can government employees use a personal ChatGPT account for work?

In most of the policies we read, no. Agencies differ on how they say it:

• GSA blocks it at the network. Its AI directive, CIO 2185.1C, signed March 11, 2026, says publicly available AI tools "not currently sanctioned for use will be blocked from the GSA network and GFE devices," and that "for chat bot use cases, users are required to use USAi." Federal nonpublic information, including "conversations that are meant to be pre-decisional or internal to GSA," can't go into any AI system without approval (GSA).
• The CPSC allows public tools, with limits. Its Generative AI Use Policy of July 8, 2025 names "ChatGPT (OpenAI), Microsoft Copilot (personal), Google Gemini, Anthropic Claude" as publicly available tools that "may transmit data across the public internet, store user input indefinitely, or use submitted content to retrain underlying models." Staff using them must not enter PII, "pre-decisional information," or confidential internal communications (CPSC).
• Pennsylvania requires state credentials. Its AI policy, effective January 13, 2026, says employees "may only use a Commonwealth approved Enterprise generative AI tool," signed in with Commonwealth credentials, and that "Personal use of non-enterprise generative AI tools with a Commonwealth e-mail is prohibited." Nothing exempt from the state's Right-to-Know Law may go into "any generative AI prompt, tool, or system" (Pennsylvania).
• Iowa says it in one line. "Personal accounts utilizing AI tools and applications cannot be used for work-related purposes" (Iowa Department of Management, effective March 31, 2025).

The ethics rules apply whichever account you use. Federal employees may not allow "the improper use of nonpublic information to further their own private interests or those of another," and nonpublic information includes anything "designated as confidential by an agency" (5 CFR 2635.703). Classified information and CUI carry their own handling rules on top of these.

Are government employees' ChatGPT chats public records?

Sometimes, and the answer depends on whether you're federal or state, and which state.

For federal employees, the National Archives set a narrow test on August 21, 2026. Its memo says that "agencies' use of AI platforms does not, in and of itself, result in the creation of federal records." Content "passively retained by a third-party platform such as ChatGPT or Claude is not necessarily 'received' by an agency, unless it is downloaded or otherwise captured in an agency system and used for official purposes." Prompts are likely records when they're "captured and saved within an agency system" and circulated or used for official purposes "other than personal convenience" (NARA). Research you never use is likely a non-record.

The federal rule on personal email doesn't obviously reach chatbots. The Federal Records Act requires an employee who creates a record in a "non-official electronic messaging account" to copy or forward it to an official account within 20 days, and an intentional violation can lead to discipline (44 U.S.C. 2911). But the law defines electronic messages as systems "used for purposes of communicating between individuals," and NARA's AI memo doesn't mention the section at all.

Washington State goes the other way. Its State Archives says that if an AI interaction, "input and output," relates to public business, it's a public record, and "Agency work-related generative AI interactions created or received using a personal device/account are still public records" (Washington State Archives, updated September 2026). Washington's Public Records Act is how reporters obtained city officials' ChatGPT logs in 2025 (Reason).

In Pennsylvania, the Office of Open Records mostly sided with the governor. When WITF asked more than two dozen state agencies for employees' ChatGPT logs, the governor's office argued they were "notes and working papers used solely for that official's or employee's own personal use," and in January 2026 the office largely agreed, WITF reports. WITF received one log, from the state's 2024 pilot with OpenAI (WITF). The Office of Open Records' own site wouldn't load for us on October 31, so we haven't read the final determination itself. No court had yet published a decision applying FOIA or a state records law to officials' AI chats as of September 22, 2026, according to an article by Steve Zansberg (Reason).

Personal AI accounts and AI records: the rules side by side

Every cell comes from the government's own page, read October 31, 2026.

In Washington, one work question in your personal ChatGPT makes that chat a public record. Federally, the same chat may never become one unless you save or use it. Either way, a work question asked in the agency's tool stays out of your personal account.

Can your agency see what you type into its AI tools?

Assume it can. NARA's memo says agencies "may decide to maintain an audit trail of all of its employees' actions on their laptops and other work devices, including actions performed in an AI system," and gives the example of a security office that captures one employee's AI history for an investigation, at which point "this audit trail has become a federal record" (NARA). What an enterprise ChatGPT, Claude, or Copilot workspace lets an administrator export is in our guide to what your employer can see on a work laptop.

An agency AI tool has already turned up in a Hatch Act case. On August 20, 2026, the Office of Special Counsel charged a VA employee with Hatch Act violations, alleging the employee used "agency resources" including "their official VA email and government-licensed Microsoft Copilot" to make campaign flyers while on duty or in the federal workplace (OSC). That's a complaint before the Merit Systems Protection Board, not a finding.

Can government employees use ChatGPT for personal use?

Yes, on their own devices and accounts. None of the policies we read reaches what you ask a chatbot on your own phone about your own life, though Pennsylvania's adds one condition: don't sign up for a personal AI tool with your Commonwealth email. The rule that follows you everywhere is the one on nonpublic information. The DeepSeek ban covers government systems. The DHS order published October 20, 2026 removes DeepSeek products and "third-party products and services developed using DeepSeek implementations" from "civilian agency information technology" (Federal Register). What that means for DeepSeek's model on other hosts is in DeepSeek alternatives.

A personal ChatGPT account still keeps your questions on OpenAI's servers, tied to your email. That's a separate privacy question from your agency's rules, and what AI chatbots collect covers it app by app.

How Morse keeps your own questions on your own device

The personal side of the problem starts with where a consumer chatbot keeps your chats: on the company's servers, under your name, for as long as its policy allows, reachable by a court order like any other stored record (can AI chats be subpoenaed). Morse is built so your chat history isn't kept there.

Morse is for your own questions, on your own device. Work material belongs in the tools your agency approves, under its rules, and a work chat can be a public record wherever you type it. For the rest of your life (a health question, a budget, a job you're weighing), here is how each mode works:

• Private, for personal questions you'd rather no AI company keep. Every new chat starts here, on open-weight models like GLM and Kimi, plus Grok. Morse has zero data retention agreements with its inference, search, and other providers, so they keep nothing, and where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't.
• Anonymous, for asking GPT, Claude, or Gemini without the provider knowing it's you. Our proxy removes your account, email, and IP address and sends the message under Morse's own account. The provider may keep the message for a limited time to check for abuse, under its own policy, so leave your name, your agency, and your title out of what you write.
• E2EE, for the questions you'd show no one. Your browser checks the enclave's hardware attestation, then encrypts the message so only that enclave can read it, and Morse's servers relay ciphertext they can't read. The enclaves are run by third-party providers, not by Morse, and the protection rests on the enclave hardware and attestation working as designed.

In every mode, your chat history is stored only in your browser, encrypted, not on Morse's servers, and our servers don't keep your messages or log what you write. It locks when you sign out and takes two keys to open, one of which never leaves your device. Morse doesn't train AI on your content, sell your data, or show ads (privacy policy), and because we don't store chats, we usually have very little to give authorities. Clearing your browser or deleting your account removes your chats for good, and Morse can't recover them, so keep a copy of anything you need. If you're a social worker in a public agency, can social workers use ChatGPT covers what your ethics code asks before a client's details go in.

On your own device, send a first message from the Morse home page without an account, or start free with no card needed. End-to-end encrypted chats come with Pro, at $19 a month.