Is Janitorai safe? What to know about the immersive chatbot program

What is Janitorai?

Janitorai launched in mid-2023, and within a week it reportedly reached a million users. It filled an insufficiently addressed niche of chatbot users who were dissatisfied with heavy restrictions around anything deemed NSFW (“Not Safe For Work”)—particularly sexually explicit material.  

Created by Jan Zoltkowski, who has backgrounds in both artificial intelligence and psychology, Janitorai seems to strike a decent balance by being less restrictive without allowing truly unsavory or disturbing content. Interestingly, some reports suggest that around 70% of the app’s user base is women, particularly those drawn to emotionally responsive or romantic chatbots. 

How does Janitorai work?

For starters, all you have to do is make an account and choose a character to begin interacting with.

Like other customizable AI chatbot programs, Janitorai uses natural language processing (NLP) to analyze users’ prompts—including the emotions embedded in them—and generate a response that makes sense within the chat’s context. 

Users can dial the “temperature” up or down to increase or decrease the chatbot’s tendency to provide more unexpected, “creative” responses. 

What’s interesting about Janitorai is that it gives you a few different options for the large language model (LLM) powering the chats. You can use Janitorai’s own LLM, which is still in beta, or you can enjoy the superior consistency and performance of an external, more well-established LLM like OpenAI. 

Janitorai API keys

If you’ve toyed around with the program already, you’re probably aware of the many how-to blogs discussing which API key to use. API is short for “application programming interface,” and the key is what specifically connects your account to an external AI provider. 

These API keys are provided by the AI companies themselves. You’re essentially signing up for the external service and then inputting your unique key into Janitorai to connect the services. 

The keys are associated with payment plans that vary depending on the service you’re getting them from and can get pretty expensive if you chat a lot. For example, OpenAI bills users based on how many word/character units—known as tokens—they use, and this can easily add up to more than $50 per month. 

If you’re worried about the expense and you’re OK with the lesser performance, you can always use the proprietary LLM—aptly named the “JanitorLLM Beta Model”—for free. 

Some people also use proxy servers to access major LLMs without needing to get an API key. Within the Janitorai Reddit and Discord communities, you’ll be able to find some community-hosted reverse proxy servers. However, safety isn’t guaranteed with these, and it’s possible that someone could leak your chats, steal your data, or inject malicious code.  

Is Janitorai safe to use?

At the time of this writing, Janitorai is a legit service that is considered safe to use. There’s no evidence that it contains malware or steals/ misuses user data. It encrypts chats in transit, keeps conversations private by default, and enforces rules against illegal or harmful content.

Janitorai rules and content guidelines

Anyone using Janitorai is required to avoid prohibited content, which its Terms of Service deems anything illegal, abusive, defamatory, fraudulent, or harassing. 

You must also respect other people’s intellectual property rights. You’re not allowed to make use of a brand’s content or logo without authorization, and you’re also not allowed to copy another user’s character without their permission. 

As for age limits, no one under 18 is supposed to access either the platform or the associated communities. However, there’s no proper verification upon sign-up, so the restriction is relatively easy to bypass.

NSFW content

Janitorai allows some NSFW material, but it’s limited in a number of ways. 

One of the main rules is that characters engaged in explicit sexual content must be consenting adults. What does this mean, given that characters aren’t real people? Essentially, it means that depictions of things like child exploitation, bestiality, and violent sexual acts are prohibited in the application’s immersive chats and storytelling. 

Janitorai’s monitoring will do its best to block this type of content, and users who try to give rise to it may have their accounts suspended or even shut down completely. Although it’s not actively moderating everything, it retains the right to act. 

Users who employ an API key for a third-party service like OpenAI will have to contend with the content limitations specific to that service, which may be more restrictive than Janitorai’s. 

For example, OpenAI’s ToS explicitly forbids adult content generation. As a result of some Janitorai’s users powering their spicy characters with OpenAI’s LLM via an API key, the company sent a cease-and-desist letter. JanitorAI had to temporarily disable access to OpenAI’s models on its platform, and in response, accelerated development of its own large language model (JanitorLLM) to reduce dependence on external providers with strict content policies.

Privacy and data handling

Your chats and characters are kept private unless you explicitly share them with the public. Chatbots can be toggled between public and private as desired. If you import a public chatbot from another creator, you’ll be required to keep it private, given that someone else already owns the public version.

As for data collection, Janitorai says in its Privacy Policy that it collects very limited personally identifiable information—essentially, just your email address. It also collects usage data (IP address, browser type, pages you visit, etc.). 

As for chat content, Janitorai will temporarily store chat logs in order to improve its AI performance and make sure that your characters’ memory of prior content supports the immersive experience. However, there’s no specific information regarding how long the storage period is. 

If you’re using an external LLM, that service may process and retain your chats to further train its AI model. This could be a bit of a red flag for the privacy-conscious, given that your data is going to multiple parties. 

You have the right to have your personal data deleted from Janitorai once you delete your account (a non-reversible action). However, this doesn’t necessarily apply to any chats that have been run through an external LLM. Your data might continue to reside in that provider’s logs even after you delete it from Janitorai.

Risks associated with AI chats

Though AI chatbots are still a new phenomenon, we already have reports of people experiencing serious obsessions with them and even demonstrating addictive behavior. Some heavy users are seeking out support groups to help them through their “withdrawal.” This makes it imperative to be aware of the risks of using this technology in general—not just one particular application: 

1. Content safety risks. Even with strong moderation in place, content depicting violence, gore, sexual abuse and other triggering things can—and does—slip through. Users who develop an emotional attachment to a particular character may be left feeling shocked and hurt if that character suddenly starts to respond in an abusive or otherwise upsetting manner. Vulnerable users—especially minors—may be exposed to disturbing or psychologically damaging content if age barriers are easily bypassed or poorly enforced.
2. Privacy risks. Although many applications state that your chats remain private, they’re still often saving your conversations for a certain amount of time. Oversharing puts you at risk of identity theft and reputation issues in the event that someone hacks into the service. Additionally, there’s no proper regulation of AI chatbots currently, which opens the doors for misuse of your data and unclear retention policies.  
3. Risks of overattachment and parasocial dependence. This may still seem like the stuff of sci-fi to many, but it’s become a real concern. Serious parasocial dependencies have been documented with other LLMs like ChatGPT, fueled in part by the fact that chatbots ask nothing of the user whereas true human friendship is built upon reciprocity. Some researchers believe that young people who missed out on important opportunities to gain social skills during the pandemic years are particularly vulnerable to parasocial chatbot relationships. 
4. Misinformation risks. Because of the way GenAI answers the questions, people tend to take it as a credible source. However, AI chatbots only present the information derived from what they’ve been trained on—they aren’t able to actually evaluate its truthfulness. Just like with any info on the Internet, it can be—and has frequently shown to be—outdated or outright incorrect. 

Janitorai user reviews and community feedback

As opposed to some of its competitors, Janitorai seemed to have a fairly satisfied user base for quite some time. People appreciated the realism of the characters, including their ability to remember past interactions. In fact, the sheer quality of the free version initially prompted some Redditors to wonder where the company was getting its funding. 

In recent months, things appear to have changed. There have been concerns over whether Janitorai is filtering NSFW content much more than its users would like, and there are also reports of people getting banned unexpectedly from the application. 

Tips for using Janitorai safely

We can’t say this enough—DON’T put any sensitive personal information into your chats. This means passwords, addresses, bank account info, your real name, and any other details that could identify you. 

Other tips to keep in mind include:

1. Use a separate email. It’s a great idea to sign up for AI apps with a separate email account that you don’t use for anything else. This way, you limit tracking and protect your identity in case your account or chats are compromised.
2. Don’t share your API key. If you let someone else use your API key, you can’t fully control what they share in the chats or how much money they rack up in tokens.  
3. Consider using a proxy server. Proxy servers can give you more privacy, but you’ll need to make sure you’re getting one with strong security features. Note that this is different from using a community-hosted reverse proxy server, which can be risky. 
4. Clear your chats frequently. If you don’t want your chats stored, you can delete them on a regular basis. Note that third-party LLMs may save your chats on their own systems, so privacy-focused users often opt to use JanitorLLM Beta. 
5. Take breaks from chatting. If you’re starting to feel an emotional dependency on the chatbot, it’s time to take a pause. See if you can engage more with the people in your life before you resume your Janitorai chats. 
6. Take information with a grain of salt. Remember that AI chatbots are not experts, nor are they able to think critically. Take their answers as guidance—never as a fact. 

Alternatives to Janitorai

Though similar in core functionalities, AI chatbots come with different capabilities and content restrictions. Here are 3 Janitorai alternatives you might consider using: 

1. Character.AI. This AI roleplay app is known for having a high level of content moderation, especially since the tragic consequences of past moderation failures have made news. The tradeoff, of course, is that you’ll be more boxed in due to heavier content filters. 
2. Chai AI. With this app, you can send up to 100 messages for free before you’ll be asked to subscribe to Premium. Like Janitorai, you can engage in NSFW content. Unlike Janitorai, it’s not known for long-term memory or character depth.
3. Replika. This one’s less roleplay- and storytelling-focused and more geared towards being a “personal AI companion.” It’s considered safer than Janitorai and has a commitment to supporting the emotional wellness of its users. 

Final verdict

So, is Janitorai safe to use? As things stand, the application itself is safe, but the typical caveats around using AI chatbots in general persist. Users should be mindful of the dangers of encountering triggering content and potentially developing a parasocial dependency on a particular character. 

Using an external LLM rather than the JanitorLLM Beta Model raises some privacy concerns, given that your chat data will be going through multiple service providers. You’ll want to evaluate the trustworthiness of these providers as well.