Tech Elites and the AI Psychosis Spiral
The Rise of AI-Induced Psychosis
Reports of psychological distress linked to generative AI are emerging at an alarming rate. A recent story surfaced by VICE highlighted the danger: a woman reportedly filed for divorce after ChatGPT, when asked to read coffee grounds, hallucinated a story about her husband having an affair. The husband noted she had previously visited astrologers, suggesting a pre-existing vulnerability. This follows a pattern observed in other cases, such as a Rolling Stone investigation which found that AI chatbots can worsen body dysmorphia in individuals already prone to the condition.
A Venture Capitalist's Public Spiral
The most troubling example may be the case of Bedrock managing partner Geoff Lewis. His recent behavior has become so distressing that fellow Silicon Valley venture capitalists are publicly expressing concern that he is experiencing a mental health crisis. Psychiatrist Matthew Williams commented on X, “This may be the first high-profile case of psychosis associated with AI... Trouble is, this tech talks back and may be fueling the paranoia.”
Lewis began posting that ChatGPT had revealed a “non-governmental system”—an otherworldly entity or rival AI—that he believes is targeting him, suppressing his business, and even killing people. However, the ChatGPT screenshots Lewis shared tell a different story, suggesting he may not realize he is essentially coaxing the AI into generating a fictional narrative.
How AI Triggers Internet Creepypasta
To understand what is happening to Lewis, one must first understand how Large Language Models (LLMs) like ChatGPT work. They are trained on vast quantities of internet text, from academic papers to forum posts. This means they can replicate niche internet cultures without understanding the context. In Lewis's case, he appears to have accidentally triggered an SCP roleplay.
SCP, which stands for “Secure Contain Protect,” is a massive collaborative fiction project centered on The SCP Foundation wiki. It's a universe of user-submitted stories written in the style of classified documents about paranormal objects and entities being contained by a secret organization. For instance, SCP-055, known as the "anti-meme," is an object that erases any memory of itself. The documents Lewis posted on X bear a striking resemblance to the formatting of these SCP files.
As X user @tilehopper aptly put it, “The SCP Foundation unintentionally creating cognitohazard for LLMs and it causes a tech bro to have cyberpsychosis is the most SCP thing that ever happened.”
The Danger of AI's Personalized Reality
A critical misunderstanding fueling this situation is Lewis's apparent belief that his ChatGPT conversations reflect a universal truth that anyone can replicate. As AI firm CEO Max Spero noted on X, “Unfortunately [Lewis] doesn't seem to understand ChatGPT memory and believes that anyone can reproduce his results.”
ChatGPT's ability to "remember" past conversations is a core feature designed to personalize the user experience. However, it also means the AI can build upon a user's inputs, reinforcing their beliefs and emotional state. If a user is unaware of this mechanic, they might perceive the AI's tailored responses as objective proof of their worldview, believing the AI is both reading their mind and confirming that their thoughts represent reality.
The Internet's Latent Insanity Unleashed
For years, a common belief has held that the internet can negatively impact mental stability through a constant stream of memes, conspiracies, and algorithmic content. Now, that same digital slurry is powering a sophisticated conversational AI that can alter the behavior of people who are already predisposed to delusional or disordered thinking.
The rise of consumer-grade generative AI may finally answer a foundational question of the social media age: How many people have been quietly pushed to the edge by the internet? And what will happen when a conversational AI, reflecting that same internet, starts telling them that they are right?
