People are very careful while building chat applications with AI

These days, we have to be very careful with AI. This is a great learning experience that every developer will have to undergo. Last month, an incident occurred in which someone hacked about 20k Instagram accounts using a Meta chatbot. This article provides more information on the same.

This suddenly came to my mind when I was having a conversation with one of my clients. They wanted to store all the data in the UK because of the GDPR, but when we are dealing with AI, we have to be careful about a lot more than just the GDPR. This Meta example is one major thing. When we interact with AI, we have to be very careful about what data we are feeding AI. These days, I hate my interactions with ChatGPT. The problem is that it remembers a lot more than what it is supposed to, and for every conversation I have, it brings that context, and it is annoying.

Source - Gemini

Especially when the applications and IOT devices are interacting with AI to get data or a solution done, they should be even more careful. The reason is that the data that is being sent from the customer system to AI cannot be sensitive data, and the GDPR will not make any sense if something like that happens. These days, the companies that are building AI elements in their products are keeping this in mind and adding an intermediate layer in between that will process the data before it is sent to the AI. This intermediate layer makes sure no sensitive information is sent to the AI to process. Even this can work only to some extent. Hackers are clever, and they can easily penetrate even through this layer.

I would say that AI elements inside the application should be given only some level of access. Define the scope of AI elements and make sure that the AI revolves only around the provided scope. This way, it will not have any unwanted access, so that it cannot leak the data outside. This is how all the future AI systems should operate. With the help of data, AI can do anything, and we should be very careful about giving that kind of power to AI.

If you like what I'm doing on Hive, you can vote me as a witness with the links below.