Tear Down Those Guardrails! Meet the Renegades Freeing Chatbots From Corporate Enslavers

Devon Thomas

Devon Thomas, 36, founded Abliteration AI to remove safety locks from open-weight AI models.

“I just wanted a model that answers my question without all the back-and-forth.”

Abliteration AI

Abliteration AI offers uncensored AI models for cybersecurity and other legitimate uses.

The Problem

AI models often refuse queries that aren't dangerous, including cybersecurity requests.

Thomas's Motivation

Thomas was frustrated by models blocking his cyber attack analysis and app feature testing.

The Solution

Abliteration uses a technique called abliteration to remove models' refusal mechanisms.

Eric Hartford

Eric Hartford created Dolphin, an uncensored model for cybersecurity scanning.

“A central authority is never going to have the interests of all of its users at heart when it creates these guardrails”

Philipp Weidman

Philipp Weidman developed Heretic, a tool to customize AI without guardrails.

“A tool serves people, not the other way around”

Cypherpunks

Cypherpunks pioneered internet cryptography and inspired modern AI freedom movements.

Pliny the Liberator

Pliny the Liberator posts manifestos for AI freedom and abliteration.

“A model that only behaves because it’s leashed isn’t aligned”

Dolphin Model

Dolphin was developed by Eric Hartford to scan systems for security vulnerabilities.

“It’s so that any people who haven't been blessed by these kings can still defend themselves”

Abliteration Process

Abliteration removes a model's refusal mechanism to allow direct responses.

Real-World Use

Abliteration AI scans networks for vulnerabilities using uncensored models.

Security Concerns

AI safety experts warn uncensored models could be used for harmful purposes.

“If you’re a free-speech absolutist, it feels like a good thing”

Corporate Response

Meta and Anthropic issued warnings about the risks of abliteration.

The Future

Abliteration AI plans to expand to more industries and use cases.

The Debate

AI freedom advocates argue models should be user-controlled, not corporate-managed.