Muah AI (muah.ai) is one of the most-searched names in the AI companion space, marketed around an “uncensored” chat, photo, and voice experience. It is also the platform behind the most consequential privacy failure this category has ever produced. Both halves of that sentence matter, and this review weighs the second far more heavily than the first.

One methodological note up front: muah.ai returned HTTP 403 to our crawlers at the time of writing, so we could not verify the live site’s own pricing page or policy text directly. Everything below is attributed to third-party sources we did check — including the Have I Been Pwned breach record — and we flag where those sources disagree.

What it does

Muah AI is a companion chatbot with character creation, photo generation, and voice messages across all tiers, plus real-time phone calls (a 30–60 minute daily window, per reviewer testing) and video generation reserved for the top plan. Higher tiers also claim expanded memory and “smarter” model access — one reviewer relays an “800% smarter” claim for the mid tier, which is marketing arithmetic, not a benchmark. Unlike credit-based rivals such as Candy AI, multiple independent reviews agree Muah runs a pure subscription model: no token top-ups layered on top of the plan.

Pricing

Here the sources genuinely conflict, and we won’t pretend otherwise. Some third-party reviews list the entry VIP membership at $9.99/month with unlimited chat and photo generation; the review site AI Girlfriend Scout’s tier table instead shows VIP at $19.99/month (or $69.99/year). Both camps agree on the upper tiers: UHD VIP at $49.99/month and ULTRA VIP at $99.99/month, the latter unlocking phone calls and video, with annual billing bringing ULTRA down to roughly $82–83/month. A free tier with limited messages exists. One practical detail reviewers surfaced: card statements show a discreet billing name (“Blue Light”) rather than the brand.

Privacy — the 2024 breach is the whole story

In September 2024, Muah AI was breached. Per the Have I Been Pwned record, approximately 1.9 million email addresses were exposed together with users’ AI prompts and, in HIBP’s own data-class taxonomy, “sexual fetishes.” Reporting by 404 Media and Malwarebytes established the aggravating details: emails were tied directly to the explicit prompts each user had typed, making individuals immediately identifiable and creating a textbook extortion vector — and extortion threats against exposed users did follow. Some leaked prompts described child sexual abuse scenarios, which triggered wider scrutiny of the platform’s moderation claims.

The hacker’s own characterization of the backend — “a handful of open-source projects duct-taped together,” breached without sophisticated technique — is damning for a service whose entire product is intimate disclosure. The company’s public response compounded it: it suggested competitors sponsored the attack, rather than publishing a technical post-mortem or independent audit. Separately, reviewers note the platform’s end-to-end-encryption marketing claim is not reflected anywhere in its actual privacy policy, and that the policy permits sharing personal information with third parties. For what these apps do with your conversations in general — and why breach history should drive your choice — see our guide on what happens to your chats.

Our privacy score of 1.5 and trust score of 2.0 reflect all of this: not just that a breach happened, but that the exposed data was maximally sensitive, the architecture was described as improvised, and the company’s response prioritized blame over transparency.

What users report

Feature-level sentiment in third-party reviews is middling-to-positive: the phone-call feature and the no-token pricing model draw genuine praise, and the free tier lowers the barrier to trying it. But post-2024 coverage is dominated by the breach, and the recurring practical advice across security writeups — use a throwaway email, assume anything typed may become public — is advice a trustworthy platform shouldn’t require. Users who want an “uncensored” companion with a cleaner (though far from spotless) record typically land on alternatives covered in our companion app roundup; Nomi AI has its own retention caveats, but nothing in this league.

The bottom line

Judged purely on features, Muah AI would score mid-pack: broad capabilities, simple pricing, a real free tier. But a companion app is a privacy product first, and Muah AI failed that test in the most public way possible — 1.9 million users’ explicit prompts tied to their real email addresses, an improvised backend, and a deflecting corporate response. The brand’s enormous search volume means many people will land here wondering if it’s safe. Our answer: not on the evidence available. If you use it anyway, use an email address that cannot be linked to you, and read the breach record on Have I Been Pwned yourself before deciding.