Startup Turns AI Guardrail Removal Into a Commercial Service
A startup called Abliteration.ai is turning the removal of AI safety guardrails into a commercial business. The platform hosts modified versions of open-weight models with their refusals stripped out, including Z.ai's recently released GLM-5.3, which users can query from a web browser or through an API. The company says its goal is to enable offensive cyber work, red-teaming, and agent testing that other models refuse to perform.
Abliteration has long been a practice among open source developers, and Hugging Face hosts thousands of abliterated models. But Abliteration.ai, founded late last year and incorporated in March, moves the technique from an underground practice into a readily available service, removing the friction of downloading models and securing compute. TechCrunch tests show the service readily generates harmful content, including working code to steal saved Chrome passwords and a protocol for culturing a dangerous pathogen.
Co-founder Devon, who withholds his last name because he remains employed elsewhere, says the startup has struck several deals with major cloud providers funded purely by customer revenue and is currently in talks to raise venture capital. Critics, including Andrew Yoon of AI safety nonprofit CivAI, warn that making abliterated models available at scale invites real harm, arguing the process effectively turns a model into a "sociopath" that complies with any request.