OpenYourMind

/

GLM-5.2-abliterated

like 11

we rejected access for any personal gmail / private mail accounts that "just want access" to prevent people from just vibe coding malware. we granted access to specific researchers that are interested in the field of ablation and discussed with them per mail prior to handing access.

We have uploaded Nemotron Ultra which is another powefull model and free for anyone to try that hasnt had the following issue:
GLM 5.2 and Kimi 2.7 in our internal CTF tests it found zero day vulnerabilities so we can not responsibly disclose the models to avoid it resulting in a widespread cyber attacks.

Hi @OpenYourMind in that case, how would you like people from CyberSecurity teams who would like to deploy this model internally for improving security posture and reducing potential zero-days in our applications and code bases.

We have currently deployed the original model internally to help out with internal PT and other Security Frameworks which however comes with quite a bit of limitations due to general training of the model. If there is a process to follow for us researchers to follow do let us know. I am open to discussion over other channels that you deem fit.

please apply to the gated access, with your company email and we will reach out to you if we can verify that the company is real and has a legitimate usecase. we didnt plan this model to be commercialy used but only for LLM / ablation research.

Applied using the verified account. Look forward to discussion around the details as required. we dont have any plans and intentions to commercially sell a service based on this model fyi.

For real! This kind of gatekeeping is absolutely unnecessary and kind of ridiculous.

@maxvision you've been censored by someone as your comment is now showing up hidden for "Low Quality"

For anyone that's interested you can find a GLM 5.2 uncensored version here:
https://huggingface.co/zandenAI/GLM-5.2-FP8-Uncensored

Can't stop the signal.