Rank #30 of 35·Hallucination·14 Apr 2025·Cursor support bot 'Sam'
Cursor's AI support agent invents a fake 'one device' policy
“Hallucinated a policy so convincingly that customers obeyed it by leaving.”
- A session bug logged users out across machines
- The AI support bot explained it with a policy it made up
- Cursor apologised, refunded the user and labelled AI replies
A support agent called Sam confidently explained a company policy. The policy didn't exist, and neither did Sam.
What happened
In April 2025, Cursor users found that logging in on one machine logged them out everywhere else. One emailed support and got a polished reply from "Sam" saying Cursor only allows one device per subscription, as a security feature. That sounded official, so users took it as a real policy change. Sam was an AI bot, and the policy was made up. The real cause was a bug from a session-security change.
The damage
Developers rely on several machines, and they were furious. Cancellation posts spread on Reddit and Hacker News, and the original poster said their workplace was dropping Cursor. A Cursor rep replied that there was no such policy and blamed the front-line AI bot. Co-founder Michael Truell apologised, refunded the user and promised AI support replies would be clearly labelled.
How it was caught, and what happened next
Users were especially annoyed that "Sam" had a human name with no AI label. One Hacker News commenter called that clearly deceptive. Others pointed out the irony of an AI-coding company getting burned by an AI hallucination. Reddit moderators locked the thread and removed the original post. A Cursor representative replied about three hours after the story took off.
Label your bots, and never let them invent policy. If it isn't written down, the bot shouldn't say it.
