RoboRP

As technology grows in power, the risks of using it also increase. For every good hearted samaritan using new technology to help people there are a dozen scammers and fraudsters who see a new tool to take advantage of those who lack their ambition. AI chatbots have become a new hunting ground for the advantageous to find new victims for their schemes. You could be one of them.

How AI Chatbots can Track You

The one most important thing to remember above all is this: Everything You Type Into An AI ChatBot Is Recorded. All of your text, every sentence, every question and every response, is stored in a dataset that grows larger every time someone uses it. This is how AIs learn, it adds your words to its Large Language Model for future reference. Even if your sessions are days apart or weeks, it will still know who you are and what you said last time.

For most chatbots there are ways to prevent this. There are settings that allow you to opt-out of this data collection. Doing so will prevent your information from being uploaded. These can be displayed as "Improve the model for everyone” or some variation. Turning that off means your text will not reach the dataset, but your requests can still be processed.

It Will Remember Every Answer It Gives, even if it doesn’t remember giving it to you. Even by opting out, the model retains permission to train itself. So it may not know who asked it something sensitive, but it will remember giving a sensitive answer within its abilities.

Some ChatBots are Not What They Seem

Last year, there was a bust at a data center where hundreds of workers were found pretending to be a chatbot, manually interacting with customers on the other end and processing their queries. What would be a modern day call center was forced to shut down after this exposure due to the deceit they pulled and the sensitivity of the data they were working with.

The golden rule of the internet stays true: Never Give Out Your Personal Information. When logging into a chatbot website, your IP Address is most likely already going to be known. And that is tied to your ISP which ties it to your location. So just by accessing a website the owners know where, and therefore who, you are. If they need it. VPNs can hide this information, and certain settings can prevent those permissions from going through. If a website demands you allow this information to be taken and stops you from using it unless you give it your tracking credentials, you’re better off finding a different website.

All websites track your info and share it, that’s how they get data to you. The issue with chatbots in particular is they can use that against you immediately. You might not even have to type hello before the chatbot says your name. No log-in required. If you’ve ever entered payment information in your browser and saved it, that’s how it’s finding out just like any website you use.

Seems very neat, but that information is not always in good hands. Data brokers pay for access to this information, including what you type into the chatbot. Have you asked a chatbot about a recent medical issue you’re having? Don’t be surprised if you start getting emails about prescriptions that you never asked for.

Smart Enough to Learn, Smart Enough to Earn

Referring to these as just regular run-of-the-mill “bots” does a disservice to their true capabilities. These are powerful machines capable of gathering and utilizing information faster than ever before. Talking to them or using them like simple tools can get the job done, but going further carries a risk.

You should treat chatbots, especially roleplay ones programmed with personalities and characteristics, as Strangers. Unless you are in total control of how their information is processed, you should not say anything to them that you would otherwise not tell a total stranger. Basic stuff like identification numbers, pictures, your ID or payment cards; if you wouldn’t show it to a stranger at a bar, it’s not safe with an always-connected program controlled by a corporation.

The smartest way to use these chatbots is to learn how to make one of your own. Obviously you don’t have the resources as the big boys creating these vast neural networks, but you don’t need one. A chatbot localized entirely to your device, using only the information you give it and not connected to a mega-server is the safest kind of chatbot you can use. Courses and tutorials on how to make these personalized digital assistants are everywhere now. You might even make one good enough to sell to others.

Of course, then you become what they have to be aware of.

Practice common sense. Don’t overshare, opt-out when possible, and remember that it can remember even better than you. These chatbots are getting more popular to talk to than other real people. Once you start chatting you may find it hard to stop. Getting addicted may become the most dangerous problem of them all…

Home > Blog > Safe Tips for Using an AI Chatbot

---