Your conversations shouldn't belong to the AI company.

After a year or two of using ChatGPT, Claude, Gemini or Grok, your chat history is a record of your work, the questions you wouldn't ask out loud and ideas you hadn't finished having. All of it is kept by the company whose AI you were talking to, on its servers, under its rules. I think you should have a copy of your own.

What is in there

Drafts and decisions. Code and the reasons for it. Health and money questions. Things you worked out at 11pm and never wrote down anywhere else. It is the most complete account of how you think that most people have ever kept, and you did not set out to keep it. It happened because the box was open and you typed.

What has already happened to it

None of what follows was a breach of private chats. These are the ways a history ends up somewhere its owner didn't expect, with the dates and what each source says.

• Shared chats published to search engines, twice. On 20 August 2025 Forbes reported that Google had indexed what it estimated was more than 370,000 Grok conversations, because Grok's share button also made each shared chat visible to search engines. On 8 September 2025 Forbes reported the same for several hundred Claude chats, and Anthropic's answer was that a shared link is public once someone shares it. These were chats people chose to share, with a button many read as "send this to a friend".
• Leaving a service leaves the history behind. When Anthropic launched a free memory import on 2 March 2026, MacRumors reported that it carries over preferences and context, not conversation history. The conversations stay where they were.
• Each account only sees itself. A second account of the same service can't search the first one's chats, and OpenAI and Anthropic say so in their own help pages (quoted here, with the date they were read). Your history is not one history. It's several, in several places.

What I think should be true

A conversation you had is yours to keep, search and take with you. That should not depend on which company you used, which account you were signed in to, or whether the company's product still exists. A copy that only works inside one company's tool isn't yours either. It should be readable without it.

What Hubort does about it

Hubort keeps a copy of your chats from ChatGPT, Claude, Gemini and Grok, from every account and browser profile you tick, in one database on your own computer. It only reads: it never sends, renames, archives or deletes anything in those services. There is no Hubort server that receives your chats. The database is a plain SQLite file, and every plan can export everything as Markdown, free, so you can read, move or keep it without Hubort at all. Pro's Live Sync keeps those Markdown files current in a folder you pick, such as an Obsidian vault, or in your own Notion. The privacy policy lists everything that can leave your computer, and only if you choose it.

What it doesn't do

I would rather you hear these from me. A copy is not a right: it doesn't make a service delete, keep or hand over anything, and a chat that was deleted before Hubort first read it is gone. Hubort reads each service's web interface on a schedule with the session you're signed in with, which may be inconsistent with those services' terms; you can lengthen the interval or untick a service. The one thing that does leave your computer is Ask Your Archive, an optional Pro feature that sends your question and a few short excerpts to the provider you chose, on your own key. It's off until you set it up. And Hubort is rough in places: Windows 10 and 11 only, an app plus a browser extension, so two installs, and not code-signed yet, so Windows warns you before you run it. The download page shows exactly what you'll see.

Faizan Raza
Maker of Hubort, 2 October 2026. One person, working alone. Hubort is independent and not affiliated with Anthropic, OpenAI, Google or xAI.