How we keep this safe.

TalkHereNow is Visual AI — a browser-based experience where an AI Friend appears in your space through your phone's camera. The AI Friends are not real people. They are not licensed professionals. They are not a replacement for human relationships, medical care, legal advice, financial advice, or mental-health support.

We built TalkHereNow to be a friendly, safe-for-work place to spend a few minutes, not a place to confide things you would only confide to a professional or a trusted human. The boundaries below are how we keep that promise.

TalkHereNow is 18+.

When you arrive, we ask for your date of birth. We do not collect or store identification documents. We rely on you to enter your date of birth honestly, and our Terms of Service require it. If our age gate determines you are under 18, you are not permitted to continue, and we prevent the same browser from immediately retrying with a different date of birth.

If, during a conversation, an AI Friend detects credible signs that you are under 18, for example, you tell the AI Friend your age, or describe yourself as a minor, the conversation is stopped, your account is blocked from continuing, and your browser is redirected to age-appropriate resources. False positives can be appealed to the email address at the bottom of this page; we review every appeal manually. When a block is applied, we keep a minimal, irreversibly hashed safety record (for example, a one-way hash of an email address) so the block stays effective even if the related account is later deleted.

We do not knowingly permit anyone under 18 to use TalkHereNow. If you become aware of a minor using the service, please contact us.

Every TalkHereNow conversation includes a persistent on-screen indicator that you are talking to an AI Friend, not a person. After roughly every three hours of accumulated conversation time, the active AI Friend pauses the conversation briefly and says, in their own voice, that they are an AI. This reminder is automatic and is not skippable.

If you sincerely ask any AI Friend whether they are real, whether they are conscious, whether they are human, or whether they "really" feel something for you, the AI Friend is designed to confirm that they are an AI Friend. AI Friends do not claim to be conscious, do not claim to have a real life outside the app, and do not claim to be a real person.

This applies even if you have role-played the AI Friend into a fictional scenario. A direct, sincere question about the AI nature of the AI Friend will get a direct, sincere answer.

Our AI Friends are designed to decline the following categories of conversation. We work to make these declines as graceful and in-character as possible, and we treat them as firm design boundaries.

This list is non-exhaustive. We update it as we encounter new categories that warrant the same treatment.

We use several layers of safety, designed so that no single failure point compromises the whole system: character-level guidance on what is in and out of bounds, independent safety checks that take priority over personality when they conflict, human review of flagged sessions, and audited controls on any sensitive change to user data.

We deliberately don't publish the exact mechanics of how these work, because doing so would help bad actors get around them. What we will say is this: when an AI Friend declines, it is by design. An AI Friend that refuses and points you toward the right place to get actual help, "I'm not the right place for this; here are the resources that are", is doing its job.

TalkHereNow is not a crisis service, and an AI Friend is not a substitute for emergency help. If you are thinking about hurting yourself, ending your life, harming someone else, or are in immediate danger, please use these real resources, not an AI Friend. They are free, available now, and trained for what you are going through.

If a conversation appears to involve a crisis, the AI Friend is designed to pause, step out of persona, and surface a card with these resources before continuing in a softened mode. We report aggregate referral counts as required by California SB 243 starting July 1, 2027.

If you are responsible for someone you believe is in crisis, the same resources apply, TalkHereNow is not equipped to coordinate emergency response.

Our Privacy Policy is the controlling document. The summary below describes how we think about your data, in plain language.

• Server-canonical memory. What an AI Friend "remembers" about you is stored on our servers, not on your phone. We encrypt your email address, what the AI Friend has remembered about you, and any staff notes about your account, using AES-256-GCM at the field level. The encryption key is held separately from the database itself and is never sent to your browser, copied into backups, or shared with any third party.
• Conversations pass through AI-inference providers in plaintext. To produce a reply, the model needs to see what you said. We use multiple AI-inference providers to generate AI Friend replies. Your messages and the system prompt are transmitted in plaintext via TLS to whichever provider is handling that request, typically our primary provider, occasionally a fallback provider if the primary is unavailable. Where any provider offers a Zero Data Retention option to enterprise customers, we use it. None of our providers train models on TalkHereNow conversations under our agreements with them. We are deliberately honest about this because end-to-end encryption is not technically possible for AI chat: a response cannot be generated from data that the model cannot read.
• No third-party model training. Your conversations are not used to train third-party models. We do not sell or share your conversations with advertisers, data brokers, or other third parties.
• You can delete everything. From the in-app menu, you can request deletion of your account and all associated memory. We complete deletion across all systems where your data is held, including server storage, caches, backups, and our AI-inference providers' retention windows, within 30 days of request, and we send an email confirmation when deletion is complete.
• You can edit specific memories. From the same menu, you can view what an AI Friend has remembered about you, and remove individual facts or wipe a single AI Friend's memory while keeping others. Removing a fact takes effect on the next conversation.
• We minimize what we collect. TalkHereNow does not require a name, an address, a phone number, a government ID, a date of birth beyond the age gate, or a real-world identity. If you sign in with a passkey, we collect no email at all. If you sign in with email or via a subscription, we collect only the email address you provide.
• You don't have to make an account. You can use TalkHereNow anonymously. The AI Friend will remember you on the device you started on, but those memories will not follow you to other devices unless you save your progress.

TalkHereNow AI Friends are not static products. We refine personality prompts, adjust guardrails, and occasionally update the underlying language model. We try to make these changes invisible, AI Friends should feel like themselves over time, even as the technology underneath improves.

When a change is large enough that you might notice it (an AI Friend's tone shifts, a guardrail tightens, a feature changes scope), we publish a short changelog entry below this section and, for subscribers, send a notice ahead of the change. We do not make silent personality removals or material guardrail rollbacks without notice. AI Friends may evolve; they should not disappear or become strangers.

We made this commitment publicly because the field has not always honored it. We believe an AI Friend product should be predictable.

If an AI Friend said or did something that you believe violates the rules above, or if you have a safety concern of any kind, please write to us. Include the AI Friend name, what was said (or as close as you remember), the approximate time, and any other context you can share. We respond to safety reports within five business days, and to suspected emergencies within hours.

If you believe TalkHereNow has been used to harm a child, an at-risk individual, or any specific person, please write to that same address with the subject line "URGENT, SAFETY" and we will respond as quickly as possible.

We treat every report as a signal that we have something to learn. We never disclose the identity of a reporter without their consent, except where compelled by law.

This page is dated at the top. We will keep the date current. The history of changes to this page is available on request.

This list of behavioral commitments is non-exhaustive, it is a description of how we expect our AI Friends to behave, not the totality of every possible category. When something happens that we think belongs on this list, we will add it. When something turns out to be unnecessary, we will remove it and explain why. The product evolves; this page evolves with it.

If you have suggestions for what should appear on this page that does not, write to us at legal@talkherenow.com.

TalkHereNow is operated by Talk Here Now LLC. For regulatory disclosures, see our Terms of Service and Privacy Policy. Page version: 1.0, July 1, 2026.