Artist’s AI secret project collects hundreds of anonymous confessions

In September 2025, Olivia Tye headed to San Francisco’s Mission Dolores Park with a makeshift box, a blank notecard, and a cardboard sign with the message “Confession Box: Tell Us Your AI Secrets” scrawled in Sharpie.

She set up a DIY confessional booth on the sidewalk, hidden from view, and waited for passersby to stop and contribute to her participatory art project.

Tai thought he might be able to get some notes that mentioned common use cases, such as “I use AI to write emails.” A few hours later, as she looked through her small collection of crowdsourced messages, one message surprised her.

“I cloned my ex’s voice to have one last conversation with him. And it helped me finally move on. Recommend 10/10.”

In just three sentences, I felt that Mr. Tai knew a lot about this author. What they are going through, how they process the sadness of a breakup, and how they use technology to bring them the closure they were looking for, something that humans couldn’t or wouldn’t get.

“From that point on, I felt like I had to keep collecting these notes and keep seeing what people would say,” Tai tells CNBC Make It. Her project “What We Tell AI” began to take shape.

Tai, 26, is an artist living in San Francisco who hosts workshops exploring the relationship between humans and artificial intelligence.

She says the most popular part of her workshops is when participants write down an AI-related confession, drop the note into a bowl, and take turns reading out the anonymous secret. It was an instant hit, so Tai picked up on the enthusiasm and set out to create a new kind of confessional experience in the age of AI, inspired by PostSecret, a 2000s community email art project created by Frank Warren.

A recent Pew Research Center study found that about half of U.S. adults now report using AI chatbots such as OpenAI’s ChatGPT, Google’s Gemini, and Microsoft’s Copilot, and one in three adults will report using them by mid-2024. Approximately one in four American adults now use these tools daily.

Dates and break-up confessions are some of the most common things Tai receives, she says. Because people ask AI for help with everything from crafting Hinge messages to getting through a situation or even practicing how to ask someone out on a date.

Tai says he has learned more secrets than he expected about people who have AI boyfriends and girlfriends.

While some of the project’s AI confessions reveal how people use technology at work and school, most revolve around how we use AI to navigate our interactions with other humans. More people are turning to AI chatbots for health and relationship-related questions, according to a study by AI Observatory, a public platform that analyzed more than 24,000 annotated AI conversations from seven different platforms from 2023 to 2026.

Tai has deployed “What We Tell AI” in more than 20 locations including the Bay Area, New York City, and Denver. Her confessional booths appear not only in public places such as parks and shopping malls, but also at gatherings where AI is the theme, such as conferences and workshops. The project has collected more than 350 handwritten notes, according to the website.

Tai said he doesn’t think this collection represents how people feel about AI as a whole, but rather an “ethnographic snapshot” of a specific place at the moment.

Some confessions express pride in using AI to help them get through difficult situations, such as this author who had to repair a relationship with an ex-lover, while others express shame and guilt over relying on AI for companionship, validation, and other help.

The note from Palo Alto, California reads, “Please forgive my sins. I have talked to AI more than I have talked to all people in my entire life,” and ends with a picture of a grimacing face.

“Basically everything in school uses AI and I feel like I don’t deserve a degree,” one New York post reads.

The project has also received some messages from people who don’t use technology. “I’ve never used AI. I’m scared of contributing to our own destruction,” says a confession from San Francisco.

Some people dispute Mr. Tai’s own assumptions. In San Francisco, where she hopes to see more pro-AI tech workers, she once received a memo about people pretending to use AI at work because they felt forced to do so by their employers, even though they personally opposed the use of AI technology.

What makes AI’s confessions so fascinating? Tai says it boils down to “the inherent curiosity we have about people’s deeply private lives, and about people whose norms differ from our own.”

It’s been just a few years since ChatGPT went public in late 2022, democratizing access to chatbots with the ability to answer increasingly personal questions using languages ​​from across the internet. Since then, a lack of norms regarding how AI should be used in work, friends and family, romantic relationships, etc. has disrupted daily life.

Additionally, there is a highly personal and even secretive nature to the way humans create AI.

“Almost everyone’s interaction history with LLM is recorded in private chat logs,” Tai says.

This confession tends to show how people are using AI to overcome conflicts with other people. Tai says these are “deeply human” issues, but also a worrying sign of how people interact.

Approximately 27% of Americans use AI chatbots to discuss personal, emotional, and social topics, and users under 50 are more likely to do so, according to a recent poll by researchers at Elon University and The Washington Post. Many users say that talking to AI makes them feel less alone and that they consider the chatbots they use most often to be their friends. Nearly 4 in 10 people who use chatbots for emotional and social reasons say they have told an AI something they would not say to someone else.

“If you normally resolve this conflict by having a conversation with another human being, what happens when the AI ​​is one of the ‘people’ you talk to? What happens when all these experiences that people are having are pushed directly into the conversation with the LLM instead of being directed at another human being? And how is that a slow, slippery atrophy of these social and relational networks?” Tai says.

In her view, large-scale language models tend to yield better results the more detail you give them, such as those that might convey personal thoughts, fears, hopes, and concerns. However, she believes that relationships can suffer when people spend their time and energy sharing that intimacy with their LLM rather than the person in their life.

Socially, Tai says, this is “probably a pretty negative thing, because[personal information]goes directly to a data center instead of to your neighbors, your family, your loved ones.”

“What We Tell AI” isn’t just about getting people to air their dirty laundry about using AI, Tai says. Rather, she says the project’s goal is to normalize the conversation about how people are using AI in their lives. She says this is easier said than done, given the current discourse about the technology, with some people holding extreme views, from people who want AI nannies to raise their children to neo-Luddites who want to go back to so-called dumb phones.

But there are many people in the middle class who are “afraid to express their views because they feel they will be shamed or punished,” she says.

People have different perspectives on how they feel about AI. Although they are generally concerned about the impact of AI on creativity and relationships, they are open to using AI for data analysis. They are optimistic about the use of AI in healthcare, but pessimistic about its impact on education and employment. Pugh said more people are using it at work.

And as they become more familiar with AI, their perspective on it is also changing. According to Pew, a growing share of American adults predict that by 2026, this technology will be bad for society, not good. Young people under 30 are particularly critical of the potential for lasting impact on society.

Tai hopes that by removing the shame and secrecy around individual AI use, people will feel more in control and be more willing to look at other big questions around AI, like how to train for an LLM or where to build a data center.

“If you’re in a situation where no one is talking about[how AI is being used]because it’s embarrassing, maybe you should start talking about it,” Tai said. “That way we can be more comfortable and honest and get to the bigger abstract (problems).”