One AI Startup Builds Models That Refuse Nothing
The startup says the model keeps reasoning and coding strength while dropping safeguards that block offensive cyber and red-teaming requests.
4 Articles
4 Articles
One AI Startup Builds Models That Refuse Nothing
Abliteration's new model removes refusals from a powerful Chinese base to enable offensive cyber and red teaming tasks. While labs tighten safeguards, uncensored variants multiply on Hugging Face. The startup draws narrow ethical lines but challenges the industry's safety consensus. This tension will shape AI deployment for years.
Abliteration.ai Sells Guardrail-Free AI Models and Is Now Courting VCs
Abliteration.ai sells web and API access to guardrail-stripped versions of frontier open-weight models like Z.AI's GLM-5.3, and the startup is now in talks with VCs despite running purely on customer revenue since incorporating in March. CivAI's Andrew Yoon warns the models can be turned into cyberattack and pathogen-culturing tools with almost no identity verification required.
While large AI laboratories are discussing the introduction of stricter security restrictions, the startup Abliteration has deliberately removed them from its new model. What is behind this? read more on t3n.de
Coverage Details
Bias Distribution
- 50% of the sources lean Left, 50% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
