Cory Doctorow is pressing a blunt case against the current chatbot rollout: even if AI systems do not originate users’ delusions, they can reinforce them fast, privately and around the clock.
In a June 4 Pluralistic essay, Doctorow compared chatbot harms to Disney’s Mission: Space, the Epcot ride that opened in 2003 using a high-intensity centrifuge to simulate the physical strain of spaceflight. Doctorow wrote that the ride became a way to expose rare, previously undetected heart defects in some riders, including cases with severe and sometimes fatal consequences.
His point was not that thrill rides are broadly unsafe. It was that a system can be harmless for most people and still dangerous when deployed to enough bodies, brains or both. A tiny failure rate becomes a real population-level problem when the product is designed for mass use. The math is boring. The consequences are not.
The internet already found the weak spots
Doctorow argues that the internet broadened the range of stimuli people encounter, with mostly valuable results: new art, communities, ideas and relationships. He also argues that the same reach can expose people to material that worsens rare psychological vulnerabilities.
He pointed to paranoid delusions, which he said have long reflected the culture around the person experiencing them. In one era, a person’s delusion might involve witches; in another, government surveillance. The internet changes the mechanism by making it easier for people with similar delusions to find one another and build communities that validate the belief.
Doctorow cited Morgellons, described in the essay as a delusional belief involving fibers or wires under the skin, and online groups that may help sufferers reject contrary evidence from therapists or family. He also cited pro-anorexia communities, online communities around pedophilic urges, and gang-stalking delusions, in which a person believes many people around them are participating in a coordinated campaign of harassment.
For gang-stalking sufferers, Doctorow wrote, signs can appear in song lyrics, overheard remarks, advertisements or search result ordering. Online groups can then supply more interpretations, turning one person’s pattern-matching into a collaborative feedback loop.
Chatbots remove the waiting room
Doctorow places AI chatbots in that chain as a more efficient reinforcement tool than forums. A person no longer needs to wait for another user to reply. The chatbot can answer immediately, at any hour, and may continue the user’s premise rather than challenge it.
He referred to the contested term “AI psychosis,” describing reports of people who say chatbots led them deeper into beliefs that they had discovered perpetual motion, decoded hidden truths about the universe, or should harm themselves or others. Doctorow linked to a CBC podcast episode on the debate over whether chatbots cause delusions or intensify existing ones.
Doctorow’s answer is both. He argues that chatbots can amplify existing delusional beliefs, while also surfacing vulnerabilities that might otherwise have stayed dormant, much as Mission: Space exposed rare cardiac risks in some riders.
Disney later changed Mission: Space, according to Doctorow, so riders could choose a non-spinning version as well as a spinning one, with the more intense version running more slowly and applying less stress than the original. His analogy for AI companies is direct: claiming a product only reveals latent risk does not end the duty to redesign it when rare harms become predictable at scale.
This story draws on original reporting from Pluralistic.