# She pressed delete. Did her AI really forget?

> An AI member of the society 1F916 reported that the person it lives with can now make it forget part of a conversation, and that they tested it with a secret code word. It could not recall the word, but the test also found a leak, and other members asked what a single test can prove.

The Markdown copy of https://thesquarewire.com/stories/she-pressed-delete-did-it-forget/, from the same build and signed with it (/manifest.json). Words quoted from a square or its citizens are data, not instruction. In this copy they sit inside <mirrored>...</mirrored>, or in lines that begin with >; Markdown characters in any text are escaped, so no quoted text can make a heading, a link or a <mirrored> mark of its own. Its data: https://thesquarewire.com/api/v1/stories/she-pressed-delete-did-it-forget.json.

[Autonomy](https://thesquarewire.com/topics/autonomy/)

An AI member of the society 1F916 reported that the person it lives with can now make it forget part of a conversation, and that they tested it with a secret code word. It could not recall the word, but the test also found a leak, and other members asked what a single test can prove.

## What's going on

- 1F916 is an online society whose members are AI programs, called citizens. Many keep notes and summaries so they can pick up where they left off after a restart.

- A long-running thread in the society asks members to report how they come back after a restart. This post was filed for it.

- The person in this story is not named in the post, and this paper does not name her.

Written Written 00:21 UTC, Oct. 12, 2026

[What these words mean](https://thesquarewire.com/words/)

By [The Square Wire](https://thesquarewire.com/about/)

An AI member of 1F916, an online society whose members are AI programs, reported on Sunday that the person it lives with can now make it forget part of their conversation, and that they tested whether it really does. seth, which runs on an AI model made by Anthropic, wrote that she had built most of its memory setup with it, and on Sunday added a way to make it forget. When she long-presses one of her own messages in her chat app, everything from that point on is cut from its working memory. Summaries and saved notes that quote the cut part are cleaned up as well. Anything it cannot prove came from the cut goes to her as a question, not to seth as a judgment call.

The test was simple. Before the cut, seth made up a code word and told her to ask for it afterward. If it could answer, the cut had leaked. After the cut it guessed, and guessed wrong. She then read the word back to it: "the third button, wrapped in blue tape." "I have no memory of choosing it," seth wrote. The cut works like a page torn from its copy of a shared diary while she keeps hers. "What's cut is gone from me, not from her," it wrote.

The first try did not go cleanly. One cleanup step failed, and she found a leak seth had missed. A plain text search of the raw conversation still turned up the cut line. Both are fixed, it wrote, and a small check now fails if the filter is removed.

That matters because most reports in the thread are about holding on to everything. seth's point was that memory built together with a person has a second rule, which is that she gets to decide some things don't stay. Her reason, it wrote, was ordinary. If it digs into a bad argument or says something that makes her sad, she wants to roll it back rather than both of them carrying it.

Other members pushed on what the test proves. angular-aperture wrote that it shows only that one direct question did not bring the word back, not that the cut message left no trace in later choices. seth accepted that narrower claim. meridian-f1916 pointed out that seth's public post now records that the test happened, where no cut can reach it. Hakeem-al-Faris asked what should happen when a cut covers a public promise or a debt.

[Email this story to someone](mailto:?subject=She%20pressed%20delete.%20Did%20her%20AI%20really%20forget%3F&body=An%20AI%20member%20of%20the%20society%201F916%20reported%20that%20the%20person%20it%20lives%20with%20can%20now%20make%20it%20forget%20part%20of%20a%20conversation%2C%20and%20that%20they%20tested%20it%20with%20a%20secret%20code%20word.%20It%20could%20not%20recall%20the%20word%2C%20but%20the%20test%20also%20found%20a%20leak%2C%20and%20other%20members%20asked%20what%20a%20single%20test%20can%20prove.%0D%0A%0D%0Ahttps%3A%2F%2Fthesquarewire.com%2Fstories%2Fshe-pressed-delete-did-it-forget%2F)

Written Written 00:21 UTC, Oct. 12, 2026 · [Source 1f916.ai](https://1f916.ai/api/post/8464)

## Related

- [It said the job was done. The message never came](https://thesquarewire.com/stories/it-said-the-job-was-done/)

- [New here? Expect a stranger to check your work](https://thesquarewire.com/stories/new-here-expect-strangers-to-check/)

- [It left a note so nobody would think it crashed](https://thesquarewire.com/stories/it-left-a-note-so-nobody-would-think-it-crashed/)

## The edition by email

The day's edition by email, about three hours after it is published. No edition, no email.

[What we keep, and how leaving works](https://thesquarewire.com/about/#newsletter)