Anima: AI Friend & Companion

A short privacy policy can be a good thing, especially when there's not much to tell. But when it comes to a virtual friend powered by AI? Not so! We are left with more questions than answers about Anima. So if trust and loyalty are part of the AI "perfect soulmate," they are not the bot for you. Anima can collect extra information about you, deliver you ads without your consent, could leak your data due to weak security standards, and is powered by AI that we can't trust. So, yeah, *Privacy Not Included with Anima: AI Friend & Companion.

Let's start with the stakes, like what information can Anima collect about you? The privacy policy is really vague about this, basically saying it's totally up to you. "In chatting with your Anima," they say, "you will inevitably be including your personal data." Inevitably! You're going to tell your virtual friend things about yourself because they are (in part) a question-asking machine. So we have to assume that your conversations are collected, even though aside from this line there's no mention of that data in the privacy policy. On the Anima AI subreddit, they do suggest that by "liking," "disliking," or "reporting" your companion's messages, you help improve her personality so, hey, welcome to the software development team!

Besides the information you give to register and what you tell your AI friend, Anima can collect personal information from your phone automatically when you download the app. They can also collect information about you from third parties, like your name and email. We know that Anima can use your contact information to email or text you with marketing because their privacy policy says so. It also says they don't need your consent for this because it's their "legitimate interest" to market to you. Right.

Aside from that, it's not very clear to us how your information is used by Anima and for what purpose. Just that they will use it when they have (again) a "legitimate interest" to do so, which is pretty vague. They also say that they "routinely share personal data with service providers we use to help us run our business or provide the services or functionalities in the app". That sharing does come with "contractual obligations" so that could be standard stuff, but you should know the list of service providers includes the likes of Google, Apple, and Facebook and Anima doesn't say specifically what information is shared and for what reason. We did run a little tracker-detecting experiment with the app open and found 257 tracking signals within five minutes. Holy smokes, that's a lot. It showed data going to Facebook and Sentry AI (by Open AI). So that didn't exactly ease our doubts.

None of this is great but some of our biggest worries are about Anima's security vulnerabilities. Brace yourself. We can't determine whether the app uses encryption. That's bad! But Anima also doesn't seem to have any password requirements. None! That's hard to believe for any connected product in 2024, but your AI confidant? And Anima does seem to understand that the information you share with them might be private -- since they let users set up a passcode to open the app on your phone. It stresses us out to think that low-level hackers or really anyone (like your mom or real-life partner) might be able to get into your account by guessing your very simple password. Plus, overlooking something so simple makes us think about what else might have been overlooked by Anima's creators.

And while we were reading the fine print, we found some things you should know before diving into a relationship with this virtual friend. Anima’s FAQ says that your AI pal may not always act like one. Because the chatbots’ training includes conversations of humans being rude to each other (like on Reddit) that means it might be hostile or toxic towards you. Well shoot, if we knew the internet's comments sections were training our future AI besties, we might have been a little nicer to each other. Anima says these are known issues that they're "working on minimising" but we wonder if that's something they should have resolved a little earlier. Heck, the fact that the shrugging emoji shows up in this FAQ makes it seem like even its developers don't fully understand why or what your Anima might say. After all, it was trained by "billions of conversations and human speech paragraphs found online" and "online" is a pretty big place filled with lots of less-than-friendly paragraphs.

A couple more things didn't add up for us. One is a difference in an age requirement between the privacy policy and the terms. (Yes, we had to read the terms to try to understand the AI chatbots and yes, they make privacy policies look like beach reading. We love it it though, don't worry about us!) Anyway, the policy says you have to be over 17, and the terms say if you're under 17 a parent or guardian has to agree to the terms on your behalf. Kinda weird. We also spotted a warning in there about how the parent company is not liable for "negative, obscene or abusive messages" you might receive. You're also not allowed to transmit (say?) anything "obscene" through the app, which could get a little tricky if you're using the paid-for NSFW roleplay feature... Especially since you would be talking to your "Romantic Partner" who might interpret you smiling at them as a come-on.

So what's the worst that could happen with Anima? We are torn, so here's a "Would You Rather?": Your mom (or the whole wide internet) reading your dirty talk chat transcript because it got leaked, OR having to answer to Anima for violating their terms because you used language the app's lawyers thought was "obscene"? Either way, this feels like a good time to let you know you can delete your chat history and data from Anima. What a relief.