Abliteration AI founded by Devon Thomas removes open-model safeguards

Devon Thomas founded Abliteration AI, a company that removes safeguards and refusal mechanisms from open-weight AI models to provide unrestricted tools for cybersecurity and other applications. Thomas and fellow advocates argue that uncensored models empower users and protect against digital threats, while critics warn that stripping guardrails increases the risk of malicious misuse, such as aiding cyberattacks or bioweapon development. Despite pushback from major tech labs and safety researchers, the technique of abliteration has gained substantial popularity among developers. Open article →