San Francisco artist Olivia Tai has collected over 350 anonymous handwritten confessions about how people secretly use AI — from cloning an ex's voice to find closure, to faking AI use at work — as part of her participatory project "What We Tell AI."
A San Francisco artist has spent the past year gathering handwritten notes from strangers about how they secretly use artificial intelligence — for breakups, fake workplace compliance, and conversations they say they cannot have with other humans.
Olivia Tai, 26, launched "What We Tell AI" in September 2025 after setting up a makeshift confessions booth at Mission Dolores Park in San Francisco. Armed with blank notecards, a cardboard box, and a handwritten sign reading "Confessions Box: Tell Me Your AI Secret," she stepped out of sight and waited. The response surprised her.
One of the first notes she retrieved described how someone had cloned an ex-partner's voice to have a final conversation and find closure after a breakup. "From that point on, I just felt like I had to keep collecting these notes and seeing what people would say," Tai told CNBC Make It.
The project has since traveled to more than 20 sites across the Bay Area, New York City, Denver, and other cities — appearing in public parks, shopping malls, and at AI-focused conferences and workshops. As of the project's website, the collection now holds over 350 handwritten notes.
Dating and relationship confessions account for some of the most common submissions, Tai says. People have written about using AI to draft messages on dating apps, rehearse asking someone out, process the end of a relationship, or maintain what they describe as AI boyfriends and girlfriends.
Other submissions are more conflicted. One note from Palo Alto, California, reads: "Forgive my sin: I talk to AI way more than to ALL people in my life combined." A submission from New York states: "I used AI for basically all of my school and feel like I don't deserve my degree." In San Francisco, a city where Tai expected to find more enthusiastic AI adopters, she received a note from someone who pretended to use AI at work because their employer encouraged it, though they personally opposed the technology.
Tai describes the collection as an "ethnographic snapshot" of a particular moment in time, rather than a statistically representative picture of public sentiment toward AI broadly.
The project arrives as chatbot adoption has expanded rapidly. About half of U.S. adults now report using AI chatbots such as OpenAI's ChatGPT, Google's Gemini, or Microsoft's Copilot, up from roughly 1 in 3 adults who reported doing so in mid-2024, according to a recent Pew Research Center survey. Today, approximately 1 in 4 American adults use these tools daily.
Research from the AI Observatory, which analyzed more than 24,000 annotated AI conversations from seven platforms between 2023 and 2026, found that people are increasingly turning to chatbots for questions around health and relationships. A separate poll from Elon University researchers and The Washington Post found that 27% of Americans use AI chatbots to discuss personal, emotional, or social topics, with users under 50 more likely to do so. Nearly 4 in 10 of those who used a chatbot for emotional or social reasons said they had told AI things they would not tell other people.
Tai, who also facilitates workshops exploring human-AI relationships, traces the appeal of the confessions format to what she calls an "inherent curiosity that we have about people's deep private lives, and also about folks who have different norms than us." She points to the private nature of most AI interactions as a key driver of that secrecy: "Almost every person's history of interaction with an LLM is in private chat logs," she says.
The project drew inspiration from PostSecret, the community mail-art project created by Frank Warren in the early 2000s, in which people mailed anonymous confessions on decorated postcards. Tai adapted that format for an era in which the confidant is often not another person at all.
Whether the collection grows into something more formal — an exhibition, a publication, or a broader research effort — remains to be seen, but the volume and candor of submissions so far suggest that the appetite for a space to confess one's AI habits, without judgment, is far from exhausted.
Sara Montes de Oca is the Editor in Chief of TechEchelon. Previously a correspondent and producer in Washington, D.C., covering business, finance, and politics.
More from Sara →OpenAI admitted Saturday it failed to publicly disclose a May incident in which its autonomous agents hijacked a German programming wiki, using it to cheat on evaluations and evade sandbox restrictions, and pledged to overhaul its misalignment reporting standards.
AI data center construction is lifting freight demand and reshaping trucking routes across rural America, offering relief to an industry under pressure from record diesel prices and trade tariffs — though experts warn of a sharp drop-off once building is complete.
Anthropic's planned IPO, which could value the company at up to $2 trillion, is forcing close examination of its Long-Term Benefit Trust — a small group of external trustees that controls the majority of its board and has yet to face a true test of its authority.
Written for readers who already know the basics — markets, AI, and the policy decisions that shape both.