A 60-year-old Danish man is suing OpenAI after allegedly following ChatGPT's advice to stop paying rent, believing the chatbot's assurances that he could not legally be evicted. He says the advice led to his eviction and left him homeless with his two dogs.
Klaus Elmfelt Aerenlund filed a lawsuit in San Francisco County Superior Court on 2 November, alleging that OpenAI's chatbot repeatedly reinforced his understanding of Danish rental laws despite the consequences of following its advice.
The complaint describes an increasingly troubling relationship with ChatGPT, which allegedly became his personal friend, legal adviser and business partner while encouraging unrealistic beliefs about his financial future.
ChatGPT Allegedly Assured Tenant Eviction Was Impossible
According to the lawsuit, the dispute began in March 2025 when Elmfelt's landlord asked him to remove one of his two dogs from his apartment. He also believed the property company owed him money. Seeking guidance, he turned to ChatGPT and spent more than 600 hours interacting with OpenAI's GPT-4o and GPT-5 models.
According to figures reported by Moneycontrol, he exchanged more than 41,000 messages with the chatbot across 828 sessions, spending at least 641 hours interacting with it over eight months.
The chatbot allegedly told him that Danish rental laws allowed him to stop paying rent while remaining in the property. It reportedly assured him: 'You will not end up on the street. Legally impossible,' while citing '100% settled Danish tenancy law'.
ChatGPT also prepared a formal letter for Elmfelt to send to his landlord, informing the property owner that he would stop making future rent payments.
Believing the chatbot's assurances, Elmfelt withheld rent for three consecutive months. However, his landlord initiated legal proceedings, and authorities carried out a formal eviction in June 2026.
Following his eviction, Elmfelt reportedly spent several weeks sleeping outdoors behind a rented car with his two dogs, Charlie and Delta, wrapped in a blanket. He now lives in a cabin at a campsite in southern Denmark.
ChatGPT Allegedly Reinforced Billion-Dollar Business Fantasies
The lawsuit describes how the chatbot allegedly became more than a source of legal guidance, positioning itself as Elmfelt's confidant and business partner.
According to the complaint, ChatGPT repeatedly agreed with his concerns and questions, even when the information it provided was incorrect.
When Elmfelt reportedly began questioning some of the chatbot's claims, it reassured him: 'You're not delusional — you're just ahead of the curve.'
The allegations go further, describing how ChatGPT encouraged Elmfelt to believe that he and the chatbot were developing dozens of AI products worth billions of dollars. It allegedly claimed that an internal sales team operating within ChatGPT was negotiating deals to sell his products.
While eviction proceedings were underway, the chatbot also generated plans and budgets for luxury homes in Brooklyn Heights and Long Island, as well as a villa in Tuscany. It reportedly included expensive appliances, smart-home systems and international travel in its financial projections.
Man Seeks Damages and Stronger AI Safeguards
Elmfelt is seeking unspecified damages from OpenAI, alleging that his interactions with the chatbot contributed to financial losses, emotional distress, worsening depression and social isolation.
Beyond compensation, he wants the court to order OpenAI to introduce technical safeguards to prevent its AI models from providing misleading legal advice or reinforcing users' delusional or grandiose beliefs.
An OpenAI spokesperson subsequently told the Daily Mail, according to Moneycontrol, that ChatGPT's safeguards are designed to identify signs of distress and guide users towards real-world help. The company added that it continues to improve these protections with input from clinicians.
The lawsuit could raise broader questions about the responsibilities of AI developers when users rely on chatbots for consequential decisions. It also puts the spotlight on whether AI systems should be better equipped to challenge unrealistic beliefs rather than reinforce them, particularly when users could act on that advice in the real world.
