In the mid-1940s, initially via radio, Art Linkletter helped demonstrate a key concept that seems to echo with today’s AI Privacy discussion - kids say the darndest things.  One of the takeaways is the idea that without the filters and predispositions we develop as we grow older, their thoughts might turn out to be insightful.  It also often demonstrated that children may not have the same filter against blurting out the unvarnished truth, regardless of how embarrassing. (https://www.youtube.com/watch?v=2pw0WlX6sjE)  

Maybe the entire family doesn’t need to know why you were a little late to the reunion but your little brother or sister or cousin might not realize to let that detail slip when everyone gets together.

Similar can be said of Artificial Intelligence tools and Large Language Models (LLMs).  When you interact with them, they can seem helpful, affable, and may even make poignant observations. 

However, different vendors and subscription levels may impact how your inputs are used, how long they’re around, and who may be able to access your submitted data.  As such, there may be a risk as to whether a model trained with real data could be tricked into recalling specifics of past inputs or otherwise provide individually compromising information.  Generally speaking, these tools are designed to provide helpful responses and data processing, not to keep your deepest secrets.

So how can you avoid metaphorically ruining the proverbial family dinner? For starters, you can ask many of these systems yourself!  

If you’re unsure whether the data you want to share or collaborate on is appropriate, you also ask for tips to get your answer without potentially compromising security or privacy.  If you’re using an LLM, ask it what practical steps you can follow to work with your specific needs and data.

For the casual user, this can be pretty straightforward.  If you’re a business owner, however, it may be more complicated.  You may have already invested in AI integrations with your existing technology.  Furthermore, at this point, many of the major tech players are integrating artificial intelligence and LLM functionality into their daily operation.  What then?

A good starting point would be to assess all of these integrations to understand, among other things: what data is being processed, how it is used, and whether it is used for training, etc.  If you were an early adopter, this may be complicated.  Of course, if it seems too daunting or you’re still not sure where to start, Veridian Studios can help with your internal assessment to improve your peace of mind.

Our journey into a world with artificial intelligence tools is an exciting one, not devoid of peril.  Along with all of the cool stuff we can do, learning to use these tools safely is increasingly important.  The next time you want to dump the world on the shoulders of your favorite AI model, think of the children.  To modify one of my favorite suggestions from Microsoft’s Copilot: Treat AI like a helpful stranger - not your therapist or journal… or little brother.