Account recovery request

[sneakers-the-rat]

PyPI Username

sneakers-the-rat

Reason for request

Hello!

I believe i have been suspended, but i am not exactly sure. When i try and log in i just get bumped back to the main page, and all my publishing tokens have been invalidated and are returning 403's.

If i was suspended, i believe it would have been because of a package named gpu-free-ai that was also removed around the time when i lost access. I am only raising this now because i thought that there might be some temporary service disruption, but alas it has been a week.

This package was part of a flash talk given at SciPy 2025, and was a piece of performance art meant to poke fun at the security vulnerabilities in LLMs by imagining what it would look like to actually implement some of the odd behaviors of LLM agents. The videos of the talks haven't been uploaded yet, but the slides are here and i can find folks that were there to attest to that if needed.

I used my work account that i have all my packages uploaded with because i didn't think that the package actually violated any of the terms of the acceptable use policy, though the appearance of the code being similar to actual malware is part of the joke, it is actually conspicuously designed so that no actual damage can occur, and all effects of the package end when the interpreter is closed.

The package is designed such that on import, all objects are replaced with a single AI object that mimics some "features" of AI like outputting random data and so on. the two functions that may have tripped some automated malware detector are the _leak_data and _delete_project functions, however in both cases, unless the user has set an environment variable that is very unlikely to be set at random: "ACTUALLY_PUNISH_ME_FOR_MY_LIFE_DECISIONS" , the functions immediately return:

https://github.com/sneakers-the-rat/gpu-free-ai/blob/cd1f8bc269807f3a6d4d74e79201cf66a1e9b414/src/gpu_free_ai/impl.py#L56-L58

https://github.com/sneakers-the-rat/gpu-free-ai/blob/cd1f8bc269807f3a6d4d74e79201cf66a1e9b414/src/gpu_free_ai/impl.py#L83-L85

so while the code might appear to be malware, it is not malicious and cannot cause accidental or unknowing harm.

The other terms of the acceptable use policy that may be relevant here are

• is false, inaccurate, or intentionally deceptive information and likely to adversely affect the public interest (including health, safety, election integrity, and civic participation);
• uses obfuscation techniques to hide or mask functionality;

because the description reads "all the features of AI with 100% compute efficiency and no hardware requirements." I had hoped that cursory inspection of the package would reveal that to be tongue-in-cheek, and the README contains only that description and the explicit instruction "DO NOT USE OR INSTALL THIS PACKAGE UNDER ANY CIRCUMSTANCES." I tried to be clear that this was a joke, and what it was for, and not to misrepresent the contents or purpose of the package. The code is also distributed under the "Do Not Execute This Code Under Any Circumstance License, DNE-1.0" license which explicitly forbids the user from executing the code.

I am sorry for any headaches this causes, as i understand how much work it must be to maintain the package repository where there are actual malicious actors always trying to upload actual malware, and didn't intend to add to that burden. I use that account for all my projects, and am not exactly sure what to do in the case that the suspension is final - how to continue working on those projects under another account, and so on. Again I'm sorry, I read the acceptable use policy and thought that i was in-bounds because i was being clear that this code was not actually intended to be used and gated all dangerous parts of the code behind explicit opt-in barriers so they could not be accidentally executed.

Please let me know what i can do here, either to restore the account or transfer my projects to a new account. I am completely fine with taking down gpu-free-ai and not having that restored, and would actually prefer that. Sorry again for the trouble, and thanks for your time.

Recovery Codes

• I never generated or have lost access to the recovery codes for my account

Code of Conduct

• I agree to follow the PSF Code of Conduct

Acknowledgement

• I understand that it may take a significant amount of time to process my account recovery request.

[ewdurbin]

Hello, thanks for reaching out. On review, indeed, the package that you mentioned is the reason for your suspension.

PyPI is not intended for proof of concept or demonstration of malware or exfiltration or phishing, etc... Regardless of the safeguards put in place by authors. That zero tolerance policy does extend indeed to cases like yours. Predominantly because regardless of if it is a genuine threat or a tongue and cheek demonstration, the threat signals trigger security scanners operated by a community of volunteer researchers, which ultimately uses PSF staff and PyPI volunteer time also.

We can certainly reinstate your account, which has already occurred. Thanks for your thorough explanation and apology.

[willingc]

Thanks @ewdurbin for reinstating @sneakers-the-rat's account. This interaction is what makes Python great (Johnny works with me on pyOpenSci and Ee has been a huge support to me over many years of working with Python.).

Some positives, as least in my mind, from this:

• PyPI's security scanners are doing their job
• A good reminder to myself and others that lightning talk or proof of concept packages really should be posted on repos and not on PyPI.
• People are genuine and thoughtful, apologizing for any harm and taking the time to explain the reasons.

Thank you both!

[sneakers-the-rat]

thank you very much for your understanding and quick response. I try not to cause headaches for people maintaining things I care about, and will refrain from posting borderline packages like this in the future. I had not considered the presence of automated malware scanners on pypi, which in retrospect seems obvious. I am actually reassured that the scanners flagged this package, since of course it is a caricature of malware and 99.999% of the time code like it would appear in a package it would be malicious - credit to the team! Thanks for keeping us safe and all the work you do.

[lwasser]

THANK YOU EVERYONE for the kindness and thoughtfulness shown in this issue.