It can hurt to discover that your relationship has been discussed in detail with an AI while you were left outside the conversation. You do not need to settle whether AI is good or bad before explaining what the secrecy or distance has done to trust.

By the end, you can:

• Name the actual hurt.

• Propose fair AI boundaries.

• Reopen direct conversation.

Direct answer: Focus first on the behaviour and its impact: private material was shared, AI advice influenced the relationship, or conversation between you stopped. Ask what your partner uses AI for, agree what should remain private and require important decisions to return to the people affected. The problem may be secrecy, substitution or disclosure, not the technology itself.

Work out what happened

Maya in Bristol learns that her girlfriend uses ChatGPT to organise thoughts before difficult conversations. Names are removed, no screenshots are uploaded and Maya receives the final words directly. She feels surprised but not betrayed. Their issue is disclosure: would knowing about the drafting help both of them trust the process?

Daniel in Leeds finds that his husband has uploaded long WhatsApp exchanges, asked whether Daniel is manipulative and quoted the AI's answer during arguments. Daniel feels as though every disagreement is now two against one. Here the issue is larger: intimate data, a one-sided judge and decisions moving away from the relationship.

Do not begin with "You must never use AI." Begin with the part you can describe.

Try a first sentence that is not an accusation

Adapt this to sound like you:

I understand why a private place to think might help. What hurt was [the messages being shared / finding out afterwards / the AI answer being used as a verdict]. I want us to agree what can be put into AI, what stays between us and how important conversations come back to us.

Ask your partner what AI makes easier. They may fear conflict, struggle to find words or want endless time to explain. Understanding that need does not require accepting every method. You can support a lower-risk alternative: bullet-point notes, a minimised hypothetical, or a conversation with a trusted person or qualified professional.

AI cannot confirm that either of you is right. It cannot see tone, repair, affection, fear or events that were not supplied. Even a careful model may reflect the framing it receives. See Why Does AI Always Agree With Me About My Relationship? and agree privacy rules using Before You Paste Private Messages Into AI: A UK Privacy Checklist.

When a joint conversation may be unsafe

If you fear monitoring, retaliation, threats, coercive control or violence, do not use this script merely because it sounds fair. Seek specialist advice about a safe route. Call 999 in immediate danger. Women's Aid offers guidance on covering your tracks online; use a safer device if you believe yours is monitored. Urgent mental-health or child-safeguarding concerns also need an appropriate professional service.

Want AI help with a difficult conversation?

If you are comparing AI for difficult conversations, Chat with Caira is most useful when the disagreement has practical or legal consequences. Caira is grounded in England and Wales legal information and can read relevant WhatsApp or iMessage messages, emails, documents, screenshots, photos and spreadsheets to help form clearer questions. She cannot authenticate the material, hear the missing side, know motives or replace the conversation.

Keep originals unchanged if they may matter. Use only a relevant working copy, minimising or redacting names, numbers, addresses, children's details and unrelated third-party information where appropriate.

Disclaimer: This article provides general information and reflection tools. It is not therapy, medical advice or legal advice. AI only sees the information you give it and may be wrong. If you are in immediate danger, contact 999. For urgent mental-health or safeguarding concerns, use an appropriate current UK professional or specialist service.