Personal story
Data privacy in the age of AI
A question from a recent job interview kept me thinking; they asked me how I take privacy into account when using AI to keep information confidential. I think it’s a very interesting question because LLMs are designed in a way that gives us a false sense of privacy, and it is very easy to forget about ethics in the process.
My answer was divided into 3 practices that have worked for me to protect information while taking advantage of the benefits of AI. These are a couple of the things I do:
Keeping information anonymous
Many times, sharing context with artificial intelligence is precisely what allows it to generate less generic and more accurate responses, avoiding receiving false analyses, data, recommendations, etc. However, among the shared documents there may be sensitive or private information, so it’s highly important to replace details such as:
Names
Identity documents
Emails and phone numbers
Credentials and API keys
Financial information
and more
So, whenever I can, I prepare an internal document specifically designed for AI systems as a way to make the process safer and more usable for me and the team.
Checking what each AI tool actually does with the information I share
While keeping information anonymous is a way to reduce risks, it’s good to remember that many of these companies use the sent information for model training; they all notify you and offer options to prevent using your chats for this purpose (for example: incognito chats), allowing us, as users, to have more control before sharing.
Following my company’s AI usage policy
Each company defines its own rules about which tool to use and the correct use of it, so before deciding based on my own judgement, I make sure my team has a list of tools and instructions about the proper use of AI in the company.
These documents generally establish:
Authorized AI platforms
Data confidentiality
Intellectual property
Person responsible for reviewing the output
and more
These practices have become an active part of my process because being responsible with data is more important than ever now that these tools have become such a big part of our everyday life and how we work. The risk doesn’t disappear completely, but I always try to reduce it as much as possible.