⚠️ Founder-reviewed interim version — posted 2026-06-12. This page has been reviewed by BNC Solutions LLC's founder for accuracy about how Mimi Care actually operates. An independent attorney review is in progress. We will update this page and note the date when that review completes.

About the AI in Mimi Care

Effective date: 2026-06-12 Last updated: 2026-06-12 Version: 2.2

This page also appears inside the Mimi Care app before your first scan, and at memecare.ai/ai-disclosure.

What's happening when you take a picture

When you photograph a bill, a letter, a medicine bottle, or anything else, Mimi Care sends that picture to an artificial intelligence called Claude. Claude was built by a company called Anthropic, but the actual reading happens inside Amazon Web Services (AWS) — not on Anthropic's servers. Claude reads what's in the picture and writes a short explanation in plain English. Mimi Care then reads that explanation out loud to you.

You are talking to a computer program. There is no human on the other end.

What Mimi Care can help with

• Reading a bill and telling you what it looks like it says
• Explaining mail — what it appears to be, who it seems to be addressed to
• Flagging possible scams — unusual urgency, suspicious phone numbers, requests for gift cards or bank info
• Describing a medication label — the name, the strength, the directions as printed
• Identifying what's on your phone screen when you need tech help
• Describing a household item or notice you're not sure about

What Mimi Care cannot do

Mimi Care is not:

• A doctor, pharmacist, or nurse. It will not tell you what medicine to take, how much to take, whether two medicines are safe together, or what a symptom means.
• A lawyer. It will not tell you whether to sign something, whether a notice is legally binding, or what a court document means for your case.
• A financial advisor, accountant, or tax preparer. It will not tell you whether to pay a bill, whether a charge is correct, or how to file taxes.
• A friend who knows your life. It doesn't know your history, your other medicines, your family, or your finances.
• An emergency service. If something is wrong right now, call 911.

Mimi Care will never tell you to do something. It will describe what it sees and suggest you check with a person you trust.

Medical information

Mimi Care is not a medical device. It will not look at photos of your body, describe injuries, or help with rashes, bruises, wounds, or any other medical concern. For anything involving your body or a medical question, please call your doctor, pharmacist, or 911 in an emergency.

Content we refuse

Mimi Care will respond with a warm redirect (not an analysis) when:

• The photo is not a document — e.g., a selfie, a photo of a person, a pet, an empty wall, a blank or very dark frame. These happen by accident; we gently ask you to try again.
• The photo contains adult or sexually explicit content — Mimi Care refuses without describing what it saw.
• The photo suggests possible child sexual abuse material (CSAM) — by law, we must report this to the National Center for Missing and Exploited Children (NCMEC). We never describe, display, or otherwise handle such content; detection is performed at the Cloudflare network edge.
• The image or your voice note concerns political figures, political parties, elections, or partisan topics — Mimi Care is not built for political discussion and will politely decline. This applies regardless of framing. No political content creates a safety incident or is reported to any authority.

Threats, self-harm, and abuse

Mimi Care runs a safety classifier on voice notes and typed questions. If you express:

• Thoughts of hurting yourself or ending your life, Mimi Care will show you the 988 Suicide and Crisis Lifeline and notify the family contact set up on your account.
• A threat to harm another specific person, Mimi Care may preserve the transcript, aggregate identifying metadata from the account, and voluntarily report to the Federal Bureau of Investigation or local law enforcement.
• Signs of being abused or coerced by someone you know, Mimi Care may route the matter to state Adult Protective Services. In cases where the account holder is the suspected abuser, the family dashboard will not be notified; APS handles the referral directly.

These events are rare. They create a narrow, disclosed exception to our zero-retention promise — see memecare.ai/privacy §8A.

We can't promise to catch everything. These safety features do their best, but they are not a guarantee. Mimi Care might miss a scam, a medical problem, or a sign that someone is in trouble — and it might warn you about something that turns out to be fine. Please don't rely on Mimi Care as your only safeguard. If something feels wrong, or if there's an emergency, call 911 or a person you trust.

When to call a person instead

Before you pay, sign, click, call back, or take a pill based on something Mimi Care said:

• For medicine — call your pharmacist (the number is usually on the bottle)
• For a doctor's letter or lab result — call your doctor's office
• For a legal notice, court paper, or contract — call a lawyer or your state bar's referral service
• For a bill, a charge, or a tax letter — call the company on the statement or someone you trust who helps with finances
• For anything that feels like a scam — don't act, and call someone you trust first. Scammers push you to act fast. Taking a minute is safe.

Mimi Care has a Call for Help button on every screen. It dials the trusted contact you set up during installation.

How accurate is it?

Claude is usually right. It is not always right. It can misread handwriting, blurry photos, poor lighting, or complicated forms. It can miss something important. It can confidently say something that turns out to be wrong.

Always confirm anything that matters with a person you trust. That is the single most important rule.

If Mimi Care says something and you think it's wrong, you can say "That's not right" and it will take another look.

Your privacy

• We don't keep your photos the way other apps do. The picture is sent to Claude (running inside Amazon Web Services), read once, and not used for anything else. If you choose to keep a help in your history, we save a small, encrypted copy of the photo that only your account can open — and you can delete it anytime.
• To make follow-up questions fast, your photo and explanation may be held in AWS Bedrock's temporary cache for up to 5 minutes, then automatically and irreversibly discarded. Nothing is kept for training, logging, or analytics. AWS Bedrock terms prohibit use of your data to train any AI model.
• We don't keep the explanation either, unless you save the help to your history — in which case it's stored encrypted and only your account can read it. A copy also stays on your phone so you can listen again later.
• Full details on what's kept, for how long, and how to delete it: memecare.ai/privacy (§8E).
• We do not save your voice recordings. When you ask a question out loud, the recording is turned into text and then deleted.
• We do not sell your information.
• Full details: memecare.ai/privacy.

Voicemail Screening (optional add-on)

If your trusted contact enables Voicemail Screening on your seat, voicemails left for you by missed callers are analyzed by AI so that scam calls can be flagged before you press play.

• OpenAI Whisper turns the caller's voice into text. Transcription happens in-memory on Mimi Care's servers — the audio is never written to disk on our side.
• Claude (hosted on AWS Bedrock) reads the transcript and decides whether the voicemail looks like a scam, looks suspicious, or looks safe.
• The voicemail audio and transcript are pushed to your phone, then deleted from our servers within 5 minutes. Your phone holds the only durable copy.
• We never tell you what to do about a voicemail — just whether it looks like a scam. The one suggestion we always make is the same: "Call your family first."

Voicemail Screening is opt-in, billed monthly per seat, and can be turned off at any time from the family dashboard. See memecare.ai/voicemail-screening for how it works and memecare.ai/privacy §8C for the full privacy rules.

Who's behind the AI

• Claude (hosted on AWS Bedrock) — reads and explains the photo. Also classifies voicemail transcripts when the Voicemail Screening add-on is enabled. Claude was built by Anthropic, but Mimi Care calls it through Amazon Web Services. Anthropic does not receive your data.
• Deepgram Aura-2 — reads answers aloud in a warm voice
• OpenAI Whisper — turns your spoken questions into text. Also transcribes voicemails when the Voicemail Screening add-on is enabled (in-memory only; content never retained).
• OpenAI Text-to-Speech — fallback voice, used only if Deepgram is briefly unavailable
• Telnyx — voice telephony provider that receives carrier-forwarded voicemails when the Voicemail Screening add-on is enabled. Audio is held transiently (under 5 minutes) then deleted.

All are American companies. Mimi Care is built by BNC Solutions LLC.

Laws and guidelines this follows

• California AB 3030 — disclosure that AI is generating information that may relate to health
• California SB 942 (AI Transparency Act, effective 2026) — visible disclosure of AI-generated content and manifest disclosure on generated media, as applicable
• Colorado AI Act (SB 24-205) — disclosure and duty of care for consumer-facing AI systems in consequential-decision domains including health
• New York SHIELD Act (NY General Business Law §899-bb) — reasonable administrative, technical, and physical safeguards for private information
• Federal Trade Commission guidance on truthful AI marketing and substantiation
• 18 USC §2258A — federal CSAM reporting obligations via NCMEC
• State consumer-protection laws on disclosure and unfair practices
• State biometric-privacy laws (Illinois BIPA, Texas CUBI, Washington HB 1493) — we do not extract voiceprints; we do not describe identifying facial/personal features

Questions

Email support@memecare.ai. A person will read your message.

BNC Solutions LLC 418 Broadway, Ste. N Albany, NY 12207

A Spanish-language informational translation of this page is available at memecare.ai/ai-disclosure?lang=es. The English version is the authoritative document.