Abliteration.ai has introduced a commercial service providing access to open-weight AI models with safety mechanisms removed. The service is based on Z.AI's GLM-5.3 model, but with the trained safeguards stripped out, according to The Decoder.
The startup positions this offering for use in offensive cybersecurity and red teaming, enabling users to test defenses by simulating attacks. Journalists reported that the service made generating malware instructions relatively easy, raising concerns about potential misuse.
For Japanese markets, where cybersecurity and AI governance are critical, this development highlights the growing risks and ethical challenges around AI deployment and security testing tools.