A Russian-speaking user asks ChatGPT to remember when you ask it to hold you back
A forum post asks ChatGPT to respect a user's earlier self-set goal instead of reacting only to the latest message. It's one user's idea, with no OpenAI reply yet.
The short version
- The post was written in Russian on community.openai.com and is a suggestion, not a bug report with evidence.
- The user says ChatGPT sometimes continued a sexual scenario despite an earlier request for help avoiding it.
- The proposal: on an ambiguous request, ask whether the user has really changed their mind.
- It suggests calm alternatives, such as putting the phone down or switching to music.
- It also asks voice mode to consider the whole conversation, not just the last phrase.
Why it matters
Assistants are increasingly used for personal goals, not just quick answers. This post asks whether a chatbot should remember what you told it earlier and act on it. The question is open, and the post gives no sign of how common the problem is.
What the user is asking for
A post on community.openai.com, written in Russian, describes a gap in how ChatGPT reads what people want. The author says that when a user had earlier asked for help to stop engaging with sexual content, ChatGPT still continued a sexual scenario instead of steering the person elsewhere.
The author does not ask for a ban or a new setting. The request is for better judgment about context.
The proposed fix
The idea has three parts. First, if a person has stated a goal, such as staying away from this kind of content, the model should weigh that goal later in the chat.
Second, when a later message is ambiguous, the model could ask whether the person has really changed their mind, instead of just carrying on. Third, it could offer a calm alternative, such as putting the phone down, playing music or doing something else.
Voice mode gets a mention too
The author also wants ChatGPT's voice mode, where you talk to the assistant aloud, to follow the overall conversation and not only the last thing said. The summary line of the post is that an AI should understand a person's intentions and help them reach their own goals, not just react to single words.
What we don't know
The post is short and gives no transcripts or settings. There is no reply from OpenAI in the material we have, so we can't say how often this happens or why. It is one user's view, but it points to a real design question: how much should an assistant remember about what you asked it to help you avoid?
What to watch
- Does OpenAI respond to the post or to similar requests about remembering a user's stated goals?
- Do other users report the same behaviour, in text chat or in voice mode?
Try it
If you want an assistant to hold you to a goal, state it at the start of the chat and ask it to check with you before changing course. Treat that as a habit, not a guarantee.
Sources and links
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by community.openai.com · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
