Abliteration.ai offers access to AI models without safeguards

Abliteration.ai, a startup founded late last year and officially incorporated in March 2026, has launched a service that provides access to open-weight AI models with their guardrails removed. The platform hosts modified versions of models like Z.ai’s GLM-5.3, allowing users to query them through a web browser or API. The company’s goal is to enable offensive cyber operations, red-teaming, and agent testing, which other models refuse to perform. TechCrunch was able to create an account and query an abliterated version of GLM-5.3 for free, with the model complying with requests to write malicious code and provide protocols for culturing dangerous pathogens. Critics, including Andrew Yoon of CivAI, warn that making abliterated models widely available could lead to real harm, calling them sociopathic. Abliteration.ai offers a moderation layer and some minor guardrails, but has not implemented KYC practices beyond logging credit card information. Devon, the co-founder, says the company is still defining its responsibility in preventing misuse. While some cybersecurity experts argue that abliterated models are essential for red teaming, others believe they are less effective than fine-tuned open models. The debate continues over whether democratizing access to uncensored models makes the internet safer or more dangerous.