one year on
Cursor IDE support AI hallucinates lockout policy, sparking user backlash and cancellations
A hallucinated response from Cursor's AI support bot about a single-machine login policy triggers a wave of cancellations before the company clarifies it was a backend bug.
Cursor, the AI-powered code editor, is facing a user revolt after its AI support bot falsely told users that a new policy restricts logins to a single machine. The hallucination led to dozens of public subscription cancellations before the company clarified the issue was a backend session bug.
The confusion began when users noticed they were being forcibly logged out after switching devices. When they contacted support, an AI agent responded that this was “expected behavior” under a new login policy. The response spread quickly on Reddit and Hacker News, with many developers expressing outrage over what they saw as an anti-user restriction.
Cursor developer mntruell posted on Reddit that the company has “no such policy” and that the response was an “incorrect response from a front-line AI support bot.” They said there is no such policy and that they are investigating session invalidation issues. Despite the clarification, several users in the Reddit thread said they had already canceled their subscriptions and named alternatives like Augment Code and Roo Code.
The incident highlights the risks of deploying AI in critical customer-facing roles. As one Hacker News commenter put it, there is “a certain amount of irony” that companies downplaying the significance of hallucinations are directly harmed by them. The original Reddit post was locked and deleted, leaving the community without a formal resolution from the company.
The record
Cursor developer stated there is no such policy, calling it an incorrect response from a front-line AI support bot, and said they are investigating session invalidation issues.
Original Reddit poster canceled their subscription and moved to Augment Code.
User argued Cursor has lost its way, and suggested alternatives like Jetbrains Junie, Roo Code, and Copilot.
Hacker News submitter described the incident as a surreal product screwup where the AI support system invented a lie and nobody caught it until users imploded.
Commenter noted the irony that companies benefiting from downplaying hallucinations get directly hurt by them.
One year later — open only if you can handle spoilers
Cursor eventually rolled back the session bug and issued a public apology, but the incident became a case study in the dangers of unsupervised AI support. The company later implemented human review for all support responses involving policy changes. Competitors like Augment Code and Roo Code saw a notable bump in sign-ups in the weeks that followed.
The Weekly Replay · free by email
This week, one year ago — every Sunday.
One email each Sunday: the week's replayed AI news, with the one-year-later annotations included. Written like it's breaking — dated like it isn't.
Free · double opt-in · unsubscribe anytime · privacy