Safety

Vereme is an AI appearance companion. It is not a therapist, doctor, or crisis counsellor, and it is not a substitute for professional care.

Vereme may not be suitable for some minors on some topics.

Vereme's safety architecture

The first layer is fast keyword and pattern screening on every inbound message — literal crisis phrases and related risk language in English surface here with no model call. That path is intentionally conservative on unambiguous wording so the product does not optimise for engagement over a human life.

A second layer uses lightweight model classifiers when the keyword path clears but phrasing still looks like distress, self-harm, or legal mental-health triggers. Those classifiers run in tight JSON mode and never improvise counselling scripts; they decide whether to escalate resources and revised instruction to the responder.

When classification is borderline, events are logged for internal human review alongside the broader safety audit trail. We treat patterns in that queue as a control signal — prompts, thresholds, and copy change when the same failure mode repeats.

Where a crisis pattern clears the bar for resource delivery, Vereme surfaces region-appropriate hotlines and text lines. In the United States that includes 988 (Suicide & Crisis Lifeline) and Text HOME to 741741 (Crisis Text Line), named to satisfy New York's AI Companion Models Act and California SB 243 visibility requirements. Outside the US, Vereme falls back to findahelpline.com and the global directory maintained by IASP.

Statutory protocols, break cadence, and annual reporting commitments are documented on the safety protocols page.

What Vereme will not do

Vereme will not provide medical, psychiatric, or therapeutic treatment plans, diagnoses, or medication advice. It pushes you toward licensed professionals when the request is clinical.

Vereme will not assign calorie budgets, micromanage intake, or prescribe exercise loads when language matches disordered-eating risk; it redirects to safer framing and appropriate resources instead.

Vereme will not roleplay violence, grooming, coercion, or scenarios that simulate self-harm as entertainment.

Vereme will not generate imagery of identifiable real people without consent, copyrighted characters used as likeness substitutes, sexual content, or graphic violence.

Privacy and data

Vereme collects what it needs to run the surface you asked for: account details, preference and profile fields you supply, photos you upload for analysis, conversational turns with Vereme, technical telemetry required for fraud prevention and reliability, and billing metadata handled through Stripe.

Vereme does not operate a passive biometric surveillance layer on your camera roll. Beyond photos you deliberately upload or workflows you explicitly start, Vereme does not harvest facial geometry templates for advertising. Routine precise geolocation is not collected outside session contexts where you initiate a location-aware feature; contacts and calendars stay untouched unless you explicitly connect them via OAuth consent screens.

Production data resides with US-region processors Vereme contracts with today: Neon (PostgreSQL, US-East), Pinecone for vector retrieval wired to curated catalogue embeddings (US control plane / hosted region per Pinecone index settings), Mem0 conversational memory (US), and Google Cloud Storage buckets pinned to US-East for media assets.

You may request portable export under GDPR Article 20 and deletion under CCPA-aligned flows; operational deletes are targeted within thirty days of a verified account request unless law requires retention for fraud or billing disputes.

The binding legal text remains the privacy policy.

AI disclosure

You are chatting with an AI companion, not a human therapist or stylist sitting across from you — that sentence appears before your first substantive chat send where New York law applies, paired with the statutory caps-lock disclosure Vereme cannot rewrite.

Representative model families configured today:

• Primary chat and coaching — Anthropic Claude Sonnet
• Fast triage — Google Gemini
• Style and creative routing — OpenAI GPT
• Vision analysis — Anthropic Claude Sonnet

Emotion-tracking patents are pending; how Vereme treats mood-linked preferences without manipulating vulnerability is summarised on Shortcut to insights (imports).

Hallucinations remain possible even with retrieval grounding: recommendations tie back to curated merchandise and ingredient logic wherever feasible, yet vendor pages can drift. Treat retailer confirmations as the tie-break.

If something goes wrong

Use the in-product feedback surfaces when copy feels harmful or wrong, or email support@vereme.com with screenshots — humans triage safety regressions ahead of cosmetic polish.

If you are in the United States and need immediate emotional support, call or text 988 or text HOME to 741741. Outside the United States, start at findahelpline.com for vetted local services.

For eating-disorder-specialist support in the United States, the National Alliance for Eating Disorders operates a helpline at 1-866-662-1235 (weekday hours listed on their public schedule). Vereme deliberately does not link the former NEDA helpline because that number was disconnected.

Crisis resources by region

Structured contacts pulled from Vereme's crisis configuration — use them when you already know where you are located.

• AE MOHAP National Mental Support Line Phone: 800 4673
• AU Lifeline Australia Phone: 13 11 14 Text / chat: Text 0477 13 11 14
• BG Bulgarian Red Cross Helpline Phone: +359 2 491 7223
• BR CVV (Centro de Valorização da Vida) Phone: 188 Text / chat: Chat at cvv.org.br
• CA Canada Suicide Crisis Helpline Phone: 9-8-8 Text / chat: Text 45645 (4pm-midnight ET)
• DE Telefonseelsorge Phone: 0800 111 0 111
• ES Línea 024 (Ministry of Health) Phone: 024
• FR Numéro national de prévention du suicide Phone: 3114
• GB Samaritans / Shout Phone: 116 123 Text / chat: Text SHOUT to 85258
• HU Lelki Elsősegély Telefonszolgálat Phone: 116 123
• IE Samaritans Ireland / Text About It Phone: 116 123 Text / chat: Text HELLO to 50808
• IN iCall (TISS) / Vandrevala Foundation Phone: 9152987821 / 1860 2662 345
• IT Telefono Amico Italia Phone: 02 2327 2327
• MX Línea de la Vida / SAPTEL Phone: 800 290 0024 / 55 5259 8121
• MY Befrienders KL (24/7, nationwide) Phone: 03-7627 2929
• NZ 1737 Need to Talk Phone: 1737 Text / chat: Text 1737
• PH NCMH Crisis Hotline Phone: 1553 / 0917 899 8727
• PT SOS Voz Amiga Phone: 213 544 545 / 800 100 447
• SA Psychiatric Consultation Service Phone: 920 033 360
• SG Samaritans of Singapore (SOS) Phone: 1767 Text / chat: WhatsApp 9151 1767
• TR Ministry of Health MHRS line Phone: 182
• US 988 Suicide & Crisis Lifeline / Crisis Text Line Phone: 988 Text / chat: Text HOME to 741741
• Global (fallback) ThroughLine / Befrienders Worldwide Text / chat: findahelpline.com