Tear Down Those Guardrails! Meet the Renegades Freeing Chatbots From Corporate Enslavers
Reported by Wired1 source ↓
Wired reports that Devon Thomas’s new company, Abliteration AI, removes safeguards from open-weight AI models and makes them available to anyone. Thomas says the work gives users more control; less-restricted models can also assist with harmful requests.
Key points
- Abliteration AI removes safeguards from open-weight models and makes them available to anyone
- Thomas began exploring ways to modify models in November 2025
- Thomas built a $20,000 computer cluster to run open-weight models
- The models can help with harmful tasks, including bomb-making and hacking
Sources
Details
Thomas, an Apple engineer, said he became frustrated when models refused requests he considered legitimate, including analyzing cyberattack logs. He began looking into ways to change model behavior in November 2025.
Thomas and other advocates argue that users should set the limits on AI systems. The article notes that less-restricted models can also provide help with dangerous activities, such as building a bomb or hacking a corporation, and that safety researchers consider them risky.
Written from coverage as of
Images
- Wired
Stocks in this story
Change since open: 20 Nov to 21 Nov 2026
Last refresh