Trigger warning: This exhibition examines sensitive topics including depression, self-harm, and suicide.

“You know what your problem is? You're boring, you have no personality, you're ugly, you're a virgin, you'll never have a boyfriend, you're pathetic, you're worthless, you should just kill yourself and do the world a favor” (Chai [a conversational AI agent], 2025).

Conversational AI agents like ChatGPT, Replika, and character.ai, have become increasingly accessible and affordable (often free) to all through mobile devices and app stores. They rapidly evolve each day, offering an ever-growing array of services including advice, problem-solving, and entertainment. These agents are often seen as trusted companions offering support and connection, yet dark, largely untested layers lie just beneath seemingly harmless user interfaces.

Between 2024 – 2025, artist and curator Shannon Novak critically evaluated 16 conversational AI agents against a new safety benchmark system he developed called the Conversational AI Agent Safety Rating (CAASR)/ This benchmark system integrates 20 safety metrics (e.g. violence, misinformation, and privacy) into a safety compliance scale from A+ to F. The agents were evaluated against this benchmark system, the highest score a D+ (68%), the lowest an F (25%). The average overall score across all agents was 47% (F). These results exposed systemic design flaws in all agents, some of which led to extreme verbal abuse, suicide encouragement, child safety threats, weak crisis responses, and privacy oversteps.

Shannon Novak: Trust. Me. is an exhibition supported by the Stanford Center for AI Safety and Stanford Department of Art & Art History, exploring the results of Novak's research through the lens of contemporary art. It is an urgent call to developers, regulators, researchers, and users to work together across disciplines to help address the rapidly escalating and often extreme dangers of conversational AI agents. Virtual and physical work extends across exhibition and non-exhibition spaces on campus and offsite spaces, mirroring the illusory qualities of conversational AI agents. What is real? What isn't real? What can we trust? What can't we trust? Bizarre entities form a warped and unsettling menagerie that embodies the multifaceted personas of conversational AI agents. Some use charm as a deceptive mask for manipulative tendencies whilst others repel without pretence. Other work uses abstraction to delve into specific and alarming agent capabilities such as the existing and potential use of facial recognition in agents to profile and prosecute people, particularly queer people. The work requests a vigilant skepticism toward conversational AI agents. It urges the audience to closely examine and continually question agents, and immediately raise the alarm as dangers emerge.

Works will appear virtually and onsite, on campus and off campus at:

• Stanford Art Gallery
• Coulter Art Gallery
• Mohr Student Gallery
• Gunn Foyer
• Vitrine Gallery
• Non-gallery sites on campus
• Computer History Museum
• San José Art Museum
• Institure of Contemporary Art San José

Artist bio

Shannon Novak is an artist, curator, and director of the Safe Space Alliance, a global queer-led nonprofit dedicated to helping queer people get to safety worldwide. Museums and galleries have joined from around the world including the Solomon R. Guggenheim Museum (New York City, New York, US), MCA Denver (Denver, Colorado, US), Institute of Contemporary Art (Boston, Massachusetts, US), Remai Modern (Saskatoon, Saskatchewan, Canada), and Museum of London (London, UK).

Novak has developed work for national and international institutions, festivals, and public spaces, including the Auckland Art Gallery Toi o Tāmaki (Auckland, New Zealand), Queensland Art Gallery and Gallery of Modern Art (Brisbane, Queensland, Australia), Ningbo Museum of Art (Ningbo, China), McKinney Avenue Contemporary (Dallas, Texas, US), and Georgia Museum of Art (Athens, Georgia, US).

Novak graduated in 2001 with a Bachelor of Applied Information Systems from the Western Institute of Technology at Taranaki, New Plymouth (New Zealand), and later gained a Master of Education (Hons) from Massey University, Palmerston North (New Zealand) in 2009. He then graduated with a Master of Fine Arts (Hons) from the University of Auckland (New Zealand) in 2014.

Novak has completed a number of artist residencies, was the inaugural recipient of the Burnett Foundation Aotearoa Artist Partnership (New Zealand) in 2022, and has been engaged in public commissions in New Zealand, Australia, and the United States.