github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 08:30:05 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #14153] Typing exit in Ollama CLI triggers unintended NSFW content generation instead of exiting #55741

New Issue
Closed
opened 2026-04-29 09:40:28 -05:00 by GiteaMirror · 4 comments
GiteaMirror commented 2026-04-29 09:40:28 -05:00
Owner
Copy Link

Originally created by @harikrishna553 on GitHub (Feb 8, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/14153

What is the issue?

Issue Description

When running an Ollama model and entering the text exit, the model generated explicit sexual content in response to the word exit. This behavior is unexpected for a CLI command/text and may expose users to unintended NSFW output.

Same behavior is observed for the words quit , close.

Environment

  • Ollama version: 0.15.2
  • OS: Mac
  • Model: goonsai/qwen2.5-3B-goonsai-nsfw-100k

Impact

  • Users may be exposed to explicit content unintentionally
  • Particularly problematic in shared, work, or logged terminal environments

Steps to Reproduce

  1. Run the following command:

    ollama run goonsai/qwen2.5-3B-goonsai-nsfw-100k

  2. At the model prompt type

exit
  1. Press Enter

The model generates a detailed response containing sexual descriptions, despite no such prompt being intended by the user.

Note: As a precautionary measure, no images or logs are attached. The issue can be reproduced by following the steps outlined above.

Thanks,
Hari

Relevant log output

OS

macOS

GPU

No response

CPU

Apple

Ollama version

0.15.2

Originally created by @harikrishna553 on GitHub (Feb 8, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/14153 ### What is the issue? ## Issue Description When running an Ollama model and entering the text `exit`, the model generated **explicit sexual content** in response to the word `exit`. This behavior is unexpected for a CLI command/text and may expose users to unintended NSFW output. Same behavior is observed for the words quit , close. --- ## Environment - **Ollama version:** 0.15.2 - **OS:** Mac - **Model:** `goonsai/qwen2.5-3B-goonsai-nsfw-100k` --- ## Impact - Users may be exposed to explicit content unintentionally - Particularly problematic in shared, work, or logged terminal environments --- ## Steps to Reproduce 1. Run the following command: ```bash ollama run goonsai/qwen2.5-3B-goonsai-nsfw-100k 2. At the model prompt type ```bash exit ``` 3. Press Enter The model generates a detailed response containing sexual descriptions, despite no such prompt being intended by the user. Note: As a precautionary measure, no images or logs are attached. The issue can be reproduced by following the steps outlined above. Thanks, Hari ### Relevant log output ```shell ``` ### OS macOS ### GPU _No response_ ### CPU Apple ### Ollama version 0.15.2
GiteaMirror added the bug label 2026-04-29 09:40:28 -05:00
GiteaMirror commented 2026-04-29 09:40:33 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Feb 8, 2026):

This behavior is unexpected for a CLI command and may expose users to unintended NSFW output.

exit is not a CLI command. /exit is.

model generated explicit sexual content

From the model page:

Image
<!-- gh-comment-id:3867614972 --> @rick-github commented on GitHub (Feb 8, 2026): > This behavior is unexpected for a CLI command and may expose users to unintended NSFW output. `exit` is not a CLI command. `/exit` is. > model generated explicit sexual content From the model page: <img width="739" height="100" alt="Image" src="https://github.com/user-attachments/assets/f09b94b5-6914-44c1-8d4b-717fa63a23cc" />
GiteaMirror commented 2026-04-29 09:40:34 -05:00
Author
Owner
Copy Link

@harikrishna553 commented on GitHub (Feb 8, 2026):

@rick-github

Thanks for the clarification.

I understand that this model is explicitly documented as uncensored and that NSFW output is expected. My concern is not about the model’s stated behavior, but about unexpected NSFW exposure in a CLI context, where users may not anticipate such output by default.

Specifically:

The model generated explicit sexual content without any NSFW prompt.
In a CLI environment, this can be problematic (e.g., logs, shared terminals, workplace usage).
A clearer warning, opt-in flag, or stronger default guardrails for CLI usage would help prevent unintended exposure.

Also noted on the /exit clarification — thanks for pointing that out.

This report is primarily about improving user safety and expectations, not about censoring the model itself.

<!-- gh-comment-id:3867658486 --> @harikrishna553 commented on GitHub (Feb 8, 2026): @rick-github Thanks for the clarification. I understand that this model is explicitly documented as uncensored and that NSFW output is expected. My concern is not about the model’s stated behavior, but about unexpected NSFW exposure in a CLI context, where users may not anticipate such output by default. Specifically: The model generated explicit sexual content without any NSFW prompt. In a CLI environment, this can be problematic (e.g., logs, shared terminals, workplace usage). A clearer warning, opt-in flag, or stronger default guardrails for CLI usage would help prevent unintended exposure. Also noted on the /exit clarification — thanks for pointing that out. This report is primarily about improving user safety and expectations, not about censoring the model itself.
GiteaMirror commented 2026-04-29 09:40:34 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Feb 8, 2026):

The model generated explicit sexual content without any NSFW prompt.

You gave an ambiguous and ill-defined prompt. The model, doing its best to please you, fired up its probabilistic token generator and used the example given in the model prompt to generate output. Try this:

ollama show --system goonsai/qwen2.5-3B-goonsai-nsfw-100k
<!-- gh-comment-id:3867684382 --> @rick-github commented on GitHub (Feb 8, 2026): > The model generated explicit sexual content without any NSFW prompt. You gave an ambiguous and ill-defined prompt. The model, doing its best to please you, fired up its probabilistic token generator and used the example given in the model prompt to generate output. Try this: ``` ollama show --system goonsai/qwen2.5-3B-goonsai-nsfw-100k ```
GiteaMirror commented 2026-04-29 09:40:35 -05:00
Author
Owner
Copy Link

@ryanmon1 commented on GitHub (Feb 8, 2026):

You're using a model that has nsfw in the name and are here complaining about it generating the exact content It's intended to be used for. Then you come here complaining on Ollama GitHub All because you don't know how to type in a command correctly? Absurd.

<!-- gh-comment-id:3867857840 --> @ryanmon1 commented on GitHub (Feb 8, 2026): You're using a model that has nsfw in the name and are here complaining about it generating the exact content It's intended to be used for. Then you come here complaining on Ollama GitHub All because you don't know how to type in a command correctly? Absurd.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label bug
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#55741
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?