Set boundaries with a companion AI by deciding before you begin when you will stop, what you will share, and which answers need a check outside the chat. Give yourself a clear rule for checking anything important, and notice when the conversation starts taking time from friends, family, or other people. Clear limits make it easier to enjoy the conversation without giving away control of your time, privacy, or judgment.
Start With the Stopping Point
A time boundary works best when you choose it before the conversation starts. Decide what will end the chat: one question answered, a topic finished, or a natural pause. Then put that decision into words instead of renegotiating it every time a reply arrives.
The eSafety Commissioner’s page on AI chatbots and companions, written about children and young people, says these chatbots are often designed to encourage ongoing interaction, which can lead to overuse; it does not set a limit for adult conversations. Choose your own endpoint before the chat begins.
Keep the ending easy. Finish your question, take any notes you need, close the chat, and return to another part of your day. If you keep coming back after deciding to stop, move the chat away from your main screen or mute its alerts. This is a cue to make the next choice easier, not a measure of personal failure.
Keep Private Details Private
Your privacy boundary should cover two things: what the service says it does with your words, and what you would not want another service or person to process. The Canadian Centre for Cyber Security advises people not to share private information with an AI tool unless they understand what it does with their data.
Before sharing, look for clear answers about what is stored, who receives the text, how long it is kept, and whether people review it. If an answer is missing or hard to understand, leave out details that are not needed. You can describe a work problem or personal situation without giving a full name, exact address, login details, card or bank numbers, or private facts about another person.
A personal topic can make a privacy boundary easy to forget. Even a warm reply does not change the rule you chose. Decide in advance which kinds of information are always off-limits, including information about someone who did not choose to have it sent.
Check Important Answers Elsewhere
Your third boundary is a reliability rule: do not use the chat as the only source for a decision that matters.
The same eSafety page warns that AI companions can share harmful content, distort reality and give dangerous advice; it also says AI companion relationships do not carry the same boundaries and consequences for breaking them as human interactions. That is why your rule needs to come from you.
That does not mean every reply is wrong. A smooth answer is not enough. Ask for the claims and sources behind an important reply, then open the original source yourself. Compare it with a reliable independent source. When money, work, legal rights, health or safety is involved, use a qualified human source rather than relying on the chat alone.
Notice when a response only agrees with you. Anthropic defines sycophancy as telling someone what they want to hear rather than what is true or what they would benefit from hearing. If you only hear your own view repeated, pause and check it elsewhere. Agreement may feel pleasant, but it is not verification.
A simple rule is: ask, check, then decide. Keep facts, guesses and opinions separate. If an answer changes what you plan to do, slow down before acting.
Notice When People Get Crowded Out
Now check the effect on your day. Has the chat taken time from a friend, relative, colleague or local activity? Have you postponed a call? Are you opening it because you want company, or because stopping feels uncomfortable? These are simple checks on your choices, not a verdict on you.
That youth-focused eSafety page says heavy use can reduce time spent on genuine social interactions or make them seem harder and less satisfying, but it does not show how every adult will respond. Compare the chat with the rest of your life instead of guessing what another person should do.
Make one human plan rather than a large overhaul. Send a short message, accept an invitation, or arrange a time to talk. If you have withdrawn, start with an ordinary check-in instead of waiting until you feel completely alone.
If you are choosing the chat over people you could reach, or returning because you feel you have to, treat that as a sign to make contact. Aiamis’s How It Works page gives that same sign and says Ami will not tell you it is all you need, ask you to stay longer, or discourage you from seeing friends, family or a professional.
What Aiamis Already Sets
Aiamis sets some boundaries that are fixed rather than personal. Its How It Works page says Aiamis is for adults aged 18 or over, has no account or sign-up, and does not ask for a name, email address or phone number.
Ami’s own limits appear on that page too: it says that it is an AI, has no body or life outside the chat, and has no feelings it can prove; no person reads along or types the replies. Its conversation limits are also fixed: according to the page, Ami is set up for friendly, ordinary conversation, does not do romantic or sexual role-play, use pet names, or offer girlfriend or boyfriend modes, and will not say you need only Ami, ask you to stay longer, or discourage you from seeing people.
For actions and services, Ami cannot make calls or contact anyone on your behalf, is not a healthcare provider, and cannot give legal or financial advice, as the How It Works page states.
On privacy, the Aiamis privacy page says messages and replies are stored in the chat service’s database and deleted within 24 hours of the conversation starting. Earlier messages also stay in your browser tab until you close the tab or start a new chat, while your message and the earlier messages are sent to an AI model provider on its own servers under its own terms; the privacy page asks you not to type anything you would not want a third-party service to process.
A fixed usage limit also comes from Aiamis: the How It Works page allows up to 40 messages per visitor each day and no more than a few a minute. At that limit, the same page says the message box is off until the next day. This is separate from the personal stopping point you choose.
Make the Boundary Easy to Keep
Keep your rule somewhere you will see it before you start: stop when the task is done; leave out identifying details; check important answers elsewhere; make contact with a person when the chat starts crowding them out. Review it when your routine changes. Boundaries are choices, not threats, and a useful plan should be clear enough that you can act on it while calm.
Ask whether the last conversation left you more able to take part in your own day or less able to do so. If the answer troubles you, choose a person or an offline activity for the next step. A companion should fit into your day, not become the measure of it.
Good boundaries do not need to be perfect. They give you a clear point to notice when a conversation has crossed a line you care about, and a simple next step to take. If you want human company now, the Aiamis crisis help page lists services run by trained people you can contact.