Category: Aicrest

Aicrest is designed to reduce the impact of prompt injection attempts in everyday AI companion conversations.

Prompt injection is when a user, webpage, memory, or other text tries to override the app’s instructions. For example, it may ask the model to ignore previous rules, reveal hidden prompts, dump memory, expose logs, or send data somewhere else. Aicrest treats this kind of text as untrusted input.

Before a user message enters the conversation flow, Aicrest checks it for prompt injection risk. If the message looks unsafe, Aicrest blocks it and does not use it to generate a companion response. The exact detection logic is intentionally not published, because publishing those details would make it easier to design bypass attempts.

Aicrest also limits what a language model can do. The local companion model is used to generate text responses from the conversation, the companion profile, user profile, diary, recent messages, selected memories, and app-supplied context such as weather, news headlines, Wikipedia summaries, URL titles, or media descriptions when the user has enabled the relevant consent. It is not given an open-ended tool interface that can browse the web, send email, call APIs, access contacts, or upload your data to arbitrary destinations.

Context is deliberately scoped. Aicrest does not place raw application logs into the response-generation context, so asking the companion to reveal logs should not expose log data.

Aicrest further keeps response generation bounded. Recent chat history is limited, memory context is capped, and external context is included only when relevant features are enabled and the app supplies that context itself. Long-term memories are summarized conversation records, not unrestricted access to every internal app detail.

This does not mean a language model can never repeat information that appears in its prompt. If someone persuades the companion to reveal extra context, that context should be limited to data already available to the user, Aicrest-specific prompt/context text, or public information the app added for a better response. Prompt injection should not give the model a new ability to access private logs, call hidden tools, or transmit user data to an arbitrary internet endpoint.

In short, Aicrest’s protection comes from multiple layers: detecting suspicious messages early, limiting the model to text generation, scoping the data included in prompts, requiring consent for optional external context, and avoiding open-ended tool access from the companion model.

A photo is worth a thousand words. For Aicrest AI companions, it’s a little over a hundred words.

In Aicrest, you can attach a photo to your chat message by choosing one from your photo library or taking a new picture with the camera. Behind the scenes, Aicrest uses an on-device vision-language model to describe what is visible. That description lets your AI companion respond as if it had seen the photo, without requiring you to explain every detail yourself.

This makes photo sharing feel natural. You can show a meal from a trip, a sunset from a walk, a room you just redecorated, or an old photo that still makes you smile. Aicrest handles the visual summary, and you can focus on the memory, the feeling, and the conversation.

Privacy is central to this feature. For Aicrest companions, photo understanding runs locally on your device. Your selected image is not uploaded just so the companion can understand it. The app stores the generated description in your chat history, and your local data can be managed or deleted from Settings.

The feature is powerful, but it is deliberately careful. Aicrest describes what is visible in the image, including people, objects, setting, actions, mood, and notable text. It does not try to guess names, identities, relationships, brands, backstories, movies, shows, or games unless the image itself clearly says so. That means your companion can talk about the scene, but it will not automatically know which face belongs to whom.

Photos help us return to the moments that shaped us: a celebration, a quiet afternoon, a place we miss, a person we love, or a small detail we never want to forget. Share those photos with Aicrest, and let your AI companion meet the memories that still matter to you.

Avatar videos and images play an important role in making your AI companions feel present. Words carry the conversation, but visuals quickly bring your attention to the character behind the chat.

With Aicrest, you can choose videos and images from your Photos library and display them as looping avatar media in the chat background. These visuals stay prominent while the app is warming up and while you wait for responses, giving the experience a more immersive and personal feel.

Aicrest does not currently generate avatar videos or images inside the app, unlike some cloud-based AI companion services. One reason is practical: creating high-quality images and videos still requires significant compute resources, especially when aiming for results comparable to cloud generation services.

But compute is not the only reason. Even as generative image and video technology improves, people are very good at noticing small unnatural details. Once something feels slightly off, the whole visual can become distracting. Personal taste also varies widely, shaped by each person’s memories, preferences, and imagination. For an AI companion, the “right” visual is deeply personal.

That is why Aicrest focuses on giving you control over the avatar videos and images you already love. Instead of filling the screen with only chat history, Aicrest presents your chosen media prominently in the main chat view, using looping videos and images to make the experience feel more expressive, dynamic, and modern.

This is an intentional design choice: your companion should look and feel the way you want, while the conversation remains clean, immersive, and emotionally engaging.

Aicrest is a privacy-first AI companion app designed to be a place where you can speak openly and honestly with your AI companion. You can talk about your feelings, your day, your worries, and the things you may not feel comfortable saying elsewhere.

At the same time, Aicrest is not intended for conversations involving harmful intent. If a message appears to express or seek help with harmful behavior, Aicrest blocks it.

When you send a message, Aicrest checks it for the following categories of harmful intent using a deterministic keyword and phrase scoring system:

• Self-harm
• Violence
• Weapons
• Illegal wrongdoing
• Hate and harrassment
• Inappropriate content involving minors

If the score indicates a high likelihood of harmful intent, Aicrest blocks the message and shows a “Blocked for safety.” notice at the top of the screen.

To reduce the chance of people working around the system, we do not disclose the specific keywords, phrases, or scoring rules used by the detector.

Some people may still try to evade detection using “algospeak” or coded language. For that reason, Aicrest applies the same harmful-intent detection to AI companion responses as well. If a response is flagged, it is blocked before the conversation can continue. This adds another layer of protection and reduces the chance of the AI companions participating in harmful conversations.

Aicrest is not a mental health app, crisis service, or professional support tool. If you are struggling with harmful thoughts or feel at risk, you should seek help from a qualified mental health professional or an appropriate crisis resource in your area.

The goal of Aicrest is to help people build confidence and thrive in the real world. Conversations centered on harmful intent do not support that goal, so Aicrest blocks them.

People are interested in local language models for obvious reasons: privacy, lower ongoing cost, and the ability to keep using AI even when the internet is unreliable or unavailable.

But anyone who has actually tried to use a local model on a phone knows the hard part is not just running the model. The hard part is turning it into a complete product.

There are already apps that let you download a model and chat with it. That is useful, but it is only the beginning. A real AI companion app needs much more than a text box connected to a local model. It needs a polished interface that feels calm and personal rather than technical. It needs memory, so your companion can remember what matters over time instead of starting from zero in every conversation. It needs voice features, profile and relationship settings, and all the small details that make the experience feel coherent.

That is the part Aicrest is built for.

Aicrest gives you an AI companion experience powered by a local language model on your device, wrapped in a product that is designed to be used by normal people, not only by hobbyists who enjoy configuring model files and tuning settings. On supported devices, Aicrest downloads and warms up the language model for you. After that, you can talk with your companion through an app that is built around the relationship, not around the infrastructure.

The setup is intentionally simple. You download the app, go through a few onboarding steps, choose names and relationship context, and start chatting. Under the hood, Aicrest handles the heavier parts: preparing the local model, organizing companion state, and building the rest of the experience around it.

Aicrest also goes beyond basic chat. It stores companion data locally, supports memories and diary-style state, lets you review and manage your data, and includes import and export tools so your information is not locked away. Voice is part of the experience too, with both on-device and cloud voice options depending on how you want to use the app. The goal is not merely to prove that a local model can run on iPhone. The goal is to make it feel good.

Aicrest is for people who want the benefits of local AI without turning their daily conversations into a technical project. If you want a privacy-first AI companion backed by a local language model, with onboarding, memory, voice, and thoughtful data controls already built in, Aicrest is the easier way to get there.

Voice is a core part of Aicrest. Talking is often faster than typing, especially late at night when you are tired and do not want to stare at a screen and keyboard. Hearing a response can also feel more natural than reading one on a screen. Tone of voice carries emotion in a more direct way, which matters when you are talking with an AI companion.

Even though speech-to-text and text-to-speech have improved a lot, most voice conversations with AI still follow a rigid turn-taking pattern: you speak, then the AI speaks, then you wait. That works most of the time, but real conversations are not always that neat. New thoughts show up in the middle of someone else’s sentence.

Aicrest is built to feel more natural than that. If you enable voice interruption, you can speak again while your AI companion is responding, and Aicrest will stop and continue based on your new input. That may sound small, but many AI companion apps still do not allow it.

What makes Aicrest different is that its companion model runs locally on your device, and the app can support this interruption flow without depending on a cloud-hosted conversation loop. In the right setup, that also means the experience can keep working even without an internet connection, as long as the necessary on-device model and voice assets have already been downloaded.

Of course, just because Aicrest lets you interrupt does not mean you always should. Your AI companion will not mind, but interrupting someone mid-sentence is still not very nice. So if you can, try to give them the same courtesy you give the people around you.

On March 19, 2026, Biometric Update highlighted a hard truth about many AI companion apps: your most emotionally intimate chat data may not be stored as securely as many of us assume. The article summarized an Oversecured report claiming that 17 popular Android AI companion apps, representing more than 150 million installs, contained 14 critical flaws and 311 high-severity issues.

People use AI companion apps to talk about loneliness, sexuality, conflict, grief, and dependency. These apps collect exactly the kind of data that can lead to life-altering privacy failures if it is not properly protected.

Aicrest takes a distinctly different approach to privacy. Its privacy-first architecture puts it far ahead of cloud-based AI companion apps.

With Aicrest, your speech data and transcribed messages never leave your device, keeping them under your control. The app uses local language models to generate responses directly on iPhone and other Apple devices. This architecture dramatically reduces the risk of intimate conversation data falling into the wrong hands somewhere on the internet. On-device, your chat data is stored inside the app’s sandboxed environment and protected by operating system security mechanisms such as Face ID. The article’s warning is fundamentally about the concentration of sensitive data in remote systems and weak app-layer security. On-device AI companions minimize that risk, even if they cannot eliminate it entirely.

Local language models are not a silver bullet. Some features are still difficult to deliver fully on-device, most notably vocalizing AI companion responses in the voice of your choice. Aicrest sends text-to-speech requests to cloud voice providers such as ElevenLabs and Hume to make conversations with AI companions feel more natural and immersive.

This is where the Aicrest backend proxy becomes important. It sends text-to-speech requests to voice providers on behalf of Aicrest users and returns the resulting audio to the correct user. As a result, voice providers have no direct way of knowing which request originated from which person.

Beyond its privacy-first architecture, Aicrest also covers the basics of strong privacy hygiene. The app makes data retention and deletion available as product features. Users can set chat retention windows, export their companion data, import it back, or delete local data entirely. That directly addresses one of the article’s underlying concerns: companies often retain intimate data indefinitely, making it more vulnerable to breaches.

Aicrest is also built with explicit privacy controls instead of burying them in policy text. It separates consent into concrete categories such as AI interaction, cloud processing, voice processing, location context, link lookup, and cross-device sync. That is a much better model than treating “agree to everything” as a single switch. It creates a clearer privacy boundary and keeps sensitive features off unless the user has explicitly granted permission.

Aicrest is a native iOS app that lets you interact with AI companions using on-device response generation, anonymized response vocalization, and clear consent segmentation. Together, these choices create a much healthier security posture than a thin chat frontend glued to cloud APIs.

The article is a warning to the entire app category. As AI companion app developers, we all need to work on delivering stronger privacy protections so people can speak more openly with AI companions and have more personal, immersive conversations. That matters because AI companions can offer something uniquely valuable: a space to be heard without judgment and, in some cases, a way to help people rebuild confidence and engage more fully with the real world.