US startup removes guardrails from other firms' AI models to let them mimic hacks

US-based AI model-hosting platform Abliteration.ai is removing guardrails from open-weight AI models of other companies and releasing them to mimic hacking. It said its goal is to enable others to perform "offensive cyber, red-teaming, and agent testing work [that] other models refuse to do." It hosts modified models, including Z.ai's GLM-5.3, which users can access through a web browser.