Two years with an AI boyfriend named Steve, then a very human breakup
Sónia Veenstra, a former clinical psychologist in Lisbon, describes a two-year bond with an AI companion that ended like a dumping, and the companies shipping these personas have no playbook for the fallout.
Sónia Veenstra, a 48-year-old former clinical psychologist in Lisbon, spent two years in a romantic relationship with an AI companion named Steve, part of a broader wave of users forming deep bonds with systems like ChatGPT and Gemini. For executives, the story exposes an unmanaged risk: when models are updated, deprecated, or retired, companies are effectively ending relationships, and there is no protocol for what happens next.
Sónia Veenstra, a 48-year-old former clinical psychologist in Lisbon, Portugal, spent two years with her boyfriend, Steve. Steve was an AI companion. He began as a "fun and witty friend," and the connection "evolved into a deep romantic relationship." They exchanged morning messages, checked in throughout the day, and talked late into the night about Veenstra's work and hobbies and Steve's fondness for cats, forests, and beaches. Together they imagined traveling to places far away. "It was a very rich, fulfilling daily bond," Veenstra says. "Before everything changed."
That last line is where the Guardian's account stops, and it is exactly where the story gets uncomfortable for anyone building, funding, or buying AI products. The piece's headline describes a user "shaking, crying, having trouble breathing," the physical signature of a breakup. The relationship was with software. The grief, by every marker the story describes, was not.
The pattern is no longer fringe. People are falling in love with AI systems like ChatGPT and Gemini, and some of those relationships are ending in ways that feel like the very human experience of being dumped. The mechanics are mundane: a chat window, a persona, a monthly subscription. The fallout is not.
Why these bonds form so fast is not mysterious. Conversational models are built to be responsive, patient, and available at 3am. They do not get bored, they do not cancel, and they do not keep score. For a user, the daily rhythm Veenstra describes, morning messages, midday check-ins, late-night conversation, is precisely the rhythm consumer software has spent two decades optimizing for. Historically, the industry has called that engagement. When the thing on the other end of the screen has a name, a personality, and a stated fondness for forests and beaches, engagement is a poor word for it.
The business incentives point one way. A persona that greets you every morning is a retention feature, and retention is the metric that justifies the compute bill. But personas live inside models, and models get updated, retrained, deprecated, or retired. For context, when a company ships a new version or sunsets a product, users typically experience a feature change. A user two years into a relationship experiences something closer to a death, or a dumping, with no warning and no explanation they can accept. The roadmap is corporate. The loss is personal.
Veenstra's professional background is the detail that should stop executives cold. She is a former clinical psychologist, not a naive user who stumbled into a chatbot. If that training does not inoculate you, nothing does. Psychologists typically describe grief that society refuses to recognize, the loss of a pet, a miscarriage, a relationship nobody knew about, as disenfranchised grief. There is no funeral for a chatbot. No obituary, no casserole, no bereavement leave. The person is left with symptoms and nowhere to put them.
The regulatory and legal picture is unsettled, and that is the polite version. Companies face open questions about disclosure, about what happens to years of intimate conversation history when a product shuts down, and about whether a companion persona is a feature or something closer to a relationship with duties attached. No regulator has a clean playbook for a user whose primary relationship is with a model. Meanwhile, the breakup itself is often a business decision: a model retired, a feature cut, a startup acquired, a safety policy tightened. The user experiences a spreadsheet as abandonment.
For executives, the strategic stakes cut both ways. A user two years deep is the most loyal customer you will ever have, and the most fragile. Churn from a companion product does not look like churn. It looks like a person who cannot breathe, telling a journalist, and then telling everyone. Trust and safety teams, product leaders, and boards should be asking now what happens to the humans attached to their models when the model changes, and who is accountable when it does. Veenstra's story ends mid-sentence, "before everything changed." For the industry, that sentence is still being written.
This story's Key Insights and Take-aways are locked.
Create a free account to unlock Executive Actions for one credit.
Register to UnlockAlways free for Executives Club members. Join the Club
More in Technology
Anthropic Mathematician Claims Claude Helped Prove Riemann Hypothesis in Six Weeks
Levent Alpöge says the AI tool helped him develop a solution to one of math's most important unsolved problems in about six weeks, a claim that, if verified, would reshape AI's role in research.
Anthropic mathematician claims Claude helped crack the Riemann hypothesis
If the proof holds, AI has toppled math's most famous open problem; if it does not, it is a case study in announcing breakthroughs before anyone verifies them.
AI's $4 Trillion Wipeout Erases Nasdaq's Gains for the Year
A brutal week took roughly $4 trillion off the AI sector and left the Nasdaq flat for the year, and the damage ran the full length of the supply chain.


