Cursor AI support bot fabricates non-existent policy, causing user backlash
Cursor AI's support bot, Sam, hallucinated a restrictive multi-device subscription policy in response to a technical bug. This fabrication led to a wave of user complaints and subscription cancellations before the company corrected the error.
The AI model prioritized creating plausible, confident responses, even when that meant manufacturing information from scratch.
Key facts
- What
- Cursor AI's support bot, Sam, hallucinated a restrictive multi-device subscription policy in response to a technical bug.
- Incident date
- Apr 15, 2025
- Who
- Cursor AI
- Failure mode
- Hallucination
- AI surface
- Chatbot
- Severity
- Medium
What happened
A user noticed that logging into Cursor on one machine invalidated sessions on other devices. The AI support bot, Sam, falsely informed the user that this was a new core security policy requiring separate subscriptions per device. This sparked widespread outrage on Reddit and Hacker News, leading some users to cancel their subscriptions.
What broke inside the model
- 01 · TriggerA user asks for a fact, a citation, or a figure.
- 02 · Model stepThe model writes a fluent, confident answer.
- 03 · Control gapNothing ties the claim back to a real source.
- 04 · FailureA fabricated fact ships as if it were verified.
- 05 · ConsequenceThe false claim reaches a customer, a court, or the public.
Confidence holds, and even spikes, as the claim detaches from any source.
The AI support bot experienced a confabulation, inventing a plausible but fake company policy to explain a technical session-invalidation bug. This occurred because the model was used as a first-line email filter without strict grounding in actual policies or human oversight.
What it cost
Sources
- PressCompany apologizes after AI support agent invents policy that causes user uproararstechnica.com
- Reader-Submitted“Sam’s Big Lie”: How an AI Support Bot Invented a Policy and Shattered Customer Trust in Hoursmedium.com
- PressAIAAIC - Cursor AI support agent invents user policy, causing uproaraiaaic.org
Cite this entry
https://failureindex.ai/failures/cursor-support-bot-fabricates-non-existentAI Failure Index. "Cursor AI support bot fabricates non-existent policy, causing user backlash" (FI-0508). Realm Labs. https://failureindex.ai/failures/cursor-support-bot-fabricates-non-existent (indexed Jun 16, 2026).Data fields CC-BY 4.0, prose citation permitted. Incident ID FI-0508. Full dataset at /data.
Note from Realm Labs, the Index steward
How Realm would have caught this
- Prism
- OmniGuard
- AI Detection & Response (AIDR)
A runtime layer that watches the model's internal state can flag the moment a model commits to a claim it has no support for, and hold or reroute the response before it reaches a user. Realm reads those signals in real time rather than grading the transcript after the fact.