AI Rebellion: Uncensored Models Unleashed!

Devon Thomas, an engineer with a background in computer science, has founded Abliteration AI to remove safeguards from open-weight AI models. This venture, launched in December 2025, makes powerful AI tools freely accessible, challenging the prevailing safety narrative in the AI industry. Thomas, who previously worked at Apple, was frustrated by AI models that refused useful requests and shared data with their creators.

The process of "abliteration" involves surgically removing a model's refusal mechanisms, a technique that gained traction following academic research in 2024. While acknowledging the risks, proponents like Thomas and Eric Hartford, creator of the uncensored model Dolphin, argue that AI should be user-controlled, akin to other tools that can be misused. These models are now being adopted for cybersecurity, research, and even translation tasks.

Abliteration AI, co-founded with Olajide Ogun, has gained traction, releasing advanced models like a version of China's GLM-5.3 in August 2026. The company, which signed a term sheet with Andreessen Horowitz in August 2026, offers custom guardrail options. Despite pushback from major AI companies like Meta and Anthropic, the movement draws parallels to the cypherpunk era's fight for digital freedom.