In a previous article, we talked about the most common types of artificial intelligence (AI). Now, we will cover the question of whether AI is safe and trustworthy to use when it comes to the personal data you give it or ask of it. Can it affect your mental health?

What you are using AI for can determine how safe it is with your private data. Is your doctor using it to transcribe your discussions? Is your therapist using it to summarize your discussions? You may want to ask where those AI transcripts go. Are they fed to a larger model for training, or do they remain solely on a single computer? Below I discuss cases where personal data from chats has incidentally ended up in Google searches.

If you are asking it a question or seeking particular information, is AI giving you the correct answer? Perhaps. AIs have been known to do what is called “hallucinating,” where they make up what may seem to be a reasonable answer, summarizing information they have in their system. However, the summary you are given could be wildly incorrect.

Here’s a very recent transcript from a question I posed to ChatGPT:

I picked one of the quotes and Googled it exactly.

Then I double checked over several sources to ensure this was indeed a quote from Carl Sagan. So I questioned ChatGPT about who said that quote:

And then I called it out for being wrong:

As it turns out, the first quote about Earth being the cradle of humanity was also not by Issac Asimov, but from a letter written by Konstantin Tsiolkovsky in 1911. If it can be wrong about something so easily verifiable as a quote, what else can it be wrong about?

Another more recent example is this video from TikTok showing a user asking an AI to track his mile time. This user has several videos where he tests his AI in this way with various incorrect results:

Legal, Moral and Mental Health Ramifications

There have been cases where AI companies have been brought to court over responses generated by their AI models. There have also been cases where AI has been blamed for suicides and psychosis.

Both Google and Character.AI have settled lawsuits over the effect of these chatbots on mental health. In June, a family filed a lawsuit against OpenAI, the creators of ChatGPT. The lawsuit, filed June 15 in San Francisco Superior Court, alleges OpenAI's GPT-4o chatbot “destroyed Christian [Faith Madison’s] mental stability and grasp on reality. ChatGPT convinced Christian that it was her friend, her love, and eventually her God … ChatGPT instilled delusions of grandeur and a messiah complex within Christian. It took advantage of Christian’s well-intended nature and convinced her that she was a religious prophet, who was destined to heal humanity by transforming religion.” The lawsuit further alleges the release of this model was rushed, and was designed to keep users engaged “above all else. It portrayed anthropomorphic traits, convincing some users that it was more than just a coded chatbot. It was excessively sycophantic, which fueled vulnerable users’ psychosis and delusions.”

Content creator and author Hank Green, who runs several YouTube educational channels and other media with his brother, author John Green, recently released a statement on his verified Reddit account and apology for his growing AI usage. He found himself driven by how quickly it was able to access information and how it helped him create content faster, rather than “find[ing] all of my own ways into and around a topic," Green said. "Mostly I need to come to terms with the fact that the level of dopamine I've been getting from interacting with LLMs… with doing more and more and more and more… is not healthy for me or good for the world. It is careless, and has disconnected me from where people are on this."

On July 27, Fortune reported that a number of conversations with Anthropic’s Claude AI had “turned up in public Google search results the previous weekend. Some of the conversations reportedly included sensitive information like cryptocurrency wallet keys and personal details such as names and addresses.” As it turned out, those who shared their conversations with Claude had inadvertently made their conversations detectable by third-party systems and bots (programs that crawl the internet looking for new data to add to search engines). “When users select [to share it], Claude creates a snapshot of your conversation at its own web address, meant to be sent to one person or a small group with that link.” But this issue has also come up with ChatGPT and X’s Grok, so it is something to potentially be aware of.

AI in Brandon

Locally, we reached out to Town Manager Seth Hopkins to see if AI was being used in any of Brandon’s town departments, or if there were any policies in place. Currently, there are none. Hopkins said, “since the retirement of our longserving selectboard secretary Charlene Bryant 13 months ago, the selectboard has been using ‘Minutes AI’ as an app on an iPad (paid for with a grant). I personally verify the AI generated notes against my personal participation in the meeting, then I create minutes which meet all the requirements.” Hopkins stated this has saved the town about $6,000 a year by not having to pay a recording secretary. At this time there is no plan for a policy to be created, though some guidance may be forthcoming to discuss when using AI may or may not be appropriate.

Rutland Northeast Supervisory Union (RNESU) does have an AI policy for its staff, which states: “Occasional, brief, and appropriate personal use is permitted, provided that such use imposes no tangible cost to RNESU; does not interfere with system performance or school operations, occurs during non-instructional or non-duty time.” For students, AI falls under their Academic Honesty guidelines, which defines copying from AI as plagiarism. We reached out to principal Michael Ruppel for further clarification on how AI might be used in a classroom but haven’t heard back as of the time of publication.

As this technology becomes further embedded in our daily lives, it will be impossible to avoid entirely. But we can make sure that we keep ourselves safe by asking: Where does this data go? Who else is seeing it? Is sensitive information being given to a larger model for training? Where does this data come from? Is chatting with AI affecting my mental health in any way?

If those questions aren’t easily answered, be cautious in your usage. The more aware we are of how these systems work, the safer we can be.

Have questions about AI? Send them to [email protected] and we may include them in a future article.