How our AI works

What it is

A conversational AI, powered by Claude from Anthropic, with instructions written by HER AGAIN. It can help you put feelings into words, plan a hard conversation, set a boundary, or untangle a decision.

What it isn’t

• It isn’t a therapist, counsellor, doctor, lawyer, financial adviser or crisis line.
• It won’t diagnose you or anyone else, or tell you whether your partner is abusive, unfaithful or anything else.
• It won’t tell you to leave, stay, confront, forgive, or decide on any timeline.
• It won’t help monitor or track another person.
• It can be wrong. Please check anything important with a qualified person.

When you might be unsafe

Before any message reaches the AI, it is checked by a simple, transparent safety screen. If it suggests you may be in danger or thinking about harming yourself, the AI is not used: you see a fixed message written by people, along with crisis lines for where you are. This works even if the AI is unavailable and even if you’ve used today’s messages.

Your conversations

• You choose how long conversations are kept: not at all, 30 days, or one year. They’re deleted automatically after that.
• Stored conversations are encrypted and never shown in our staff tools.
• Messages are sent to Anthropic to generate a reply. Anthropic does not train its models on this data. Their retention terms are listed in our privacy notice.
• HER AGAIN never uses your conversations to train AI models. If we ever wanted to, we would ask you separately, and the answer would default to no.
• You can delete one conversation or all of them at any time.

Healthy use

The companion is meant to support, not replace, the people in your life. It will sometimes encourage you to reach out to a friend, a support group or a professional. Daily message limits are part of that design.