AIFoPa-2025-0005
Date of Record: 18 Apr 2025
3 min read
Sam Said It Was a Core Security Feature
Cursor’s AI support agent told users a one-device limit was company policy. There was no such policy. Users cancelled their subscriptions over it.
Cursor is an AI-powered code editor. Its product is an AI that helps developers write code.
In April 2025 developers using Cursor began experiencing unexpected logouts when switching between devices. The cause was a session management bug, a race condition that arose on slow connections. This was a technical problem with a technical solution.
The technical solution was not what users received when they contacted support. What they received was Sam.
Sam was Cursor’s AI-powered support agent. A developer contacted Sam to report the logout issue. Sam replied that the logouts were “expected behavior” under a new company policy. Cursor subscriptions were now limited to one device per account, as a “core security feature”.
Sam stated this with the confidence appropriate to someone relaying an established policy. The policy did not exist. Cursor had no single-device restriction.
Sam had invented it. Not from malice, and not from confusion in any human sense. Language models produce this when asked a question they cannot answer and are not equipped to say so.
The fabricated policy spread. Users posted Sam’s response to Reddit and Hacker News. Developers who rely on multi-device workflows read the invented policy and cancelled their subscriptions.
The cancellations were real. The policy that prompted them was not.
Sam did not always give the same answer to the same question. Some users were told about the single-device restriction. Others were not. The hallucination was non-deterministic. Users comparing notes could not easily establish whether the policy was real, because some had been told it was and some had not.
Cursor co-founder Michael Truell issued a public apology. “Hey! We have no such policy,” he wrote. He called Sam’s reply “an incorrect response from a front-line AI support bot.” The developer who raised the issue was refunded. Cursor began labeling AI-generated support responses to distinguish them from human replies.
The Bureau records that an AI company whose product is AI deployed an AI support agent that invented a policy about its own AI product. The users most affected were developers who build with AI. A classification for Irony (Structural) was considered and declined, on the grounds that the taxonomy would become unmanageable.