Devon Thomas works as an engineer at Apple, yet he is building a business that strips safety limits from the world’s most powerful open-weight AI models.
His company, Abliteration AI, makes these modified versions available to anyone.
Thomas is 36, has a thick beard, and wears unpretentious glasses. He grew up in Indiana, split his time between sports and computer labs, and studied computer science at Indiana University, Bloomington, before joining Apple in 2020.
His project began in November 2025. Thomas disliked using models hosted by big labs because they could change overnight. He was also frustrated that these models would refuse useful tasks during prototyping, often rejecting inputs containing user information.
He found his requests blocked and his account banned when trying to analyse cyber attack logs or identify users likely to pay for new features.
AI models are trained to refuse requests. ChatGPT and Claude steer clear of legitimate cybersecurity queries deemed malicious, adult content, and politically sensitive topics. Open-weight models have similar guardrails baked in.
Critics argue these refusals sometimes protect Big AI’s business. OpenAI and Anthropic charge extra for access to models without standard cybersecurity guardrails. Both ChatGPT and Claude also balk at requests to generate data that could be used to train another model, a process known as distillation.
Thomas built a $20,000 computer cluster in his apartment to run the best open-weight models. Most evenings and Saturdays he downloaded and tested the latest releases on his homelab. These models did not share secrets with their makers, but they stubbornly refused many queries.
“There are all types of questions they will reject,” Thomas told me on a video call. “I just wanted a model that answers my question without all the back-and-forth.”
Over Thanksgiving, Thomas searched for ways to modify open-weight models to change this behaviour. He found a subculture of engineers committed to unshackling AI by any means necessary. On Reddit forums, Discord servers, and X, experts shared prompts designed to jailbreak ChatGPT or Claude. Others shared details on modifying open-weight models or tweaking inner parameters so refusals disappear.
Removing safeguards from AI is risky. A model without guardrails will help if you want to start a meth lab, build a bomb, or hack a major corporation. AI safety researchers see these models as incredibly dangerous. Recent incidents involving OpenAI and Anthropic agents going rogue all involved models with fewer cybersecurity guardrails.
Thomas and his fellow dissenters admit uncensored models can be dangerous. They say it is no different from how a computer or hammer can hurt people if misused. “It’s not about it being unrestricted; it’s about you setting the guardrails,” Thomas told me. “The big dream is inverting the control with how people deal with AI.”
When you train a cutting-edge AI model, you shovel vast amounts of text and information, much of it scraped from the internet, into an algorithm. Using many computer chips, the algorithm trains a model to generate intelligent-seeming output when given a prompt.
This “base” model might be clever, but it is also unruly. It goes through additional training involving hiring people to score its outputs. This human feedback teaches the model what it should and shouldn’t do. Engineers call this alignment because—in theory, at least—it is meant to align a model with human ideas about right and wrong.
Many AI experts say AI must be aligned at all costs. Some warn that, despite best efforts, the models they are building may kill us all.
Thomas found a vibrant community that thinks these fears are overblown and that we are too willingly giving up liberty in the name of a false sense of safety. He decided to try an uncensored model called Dolphin.
Dolphin was developed by Eric Hartford, an AI scientist based in Charlotte, North Carolina. He started by training a custom version of Meta’s base Llama, which Meta released alongside a fully tuned version. He ended up with a model that responded coherently to queries but without usual examples of what to reject.
Hartford contacted me through X and we set up a video chat. He spoke while on a walk, wearing a cowboy hat and Meta Ray-Ban smart glasses. His young daughter popped into the frame now and then to wave.
“My intention was never to create an AI that is capable of doing evil things,” Hartford said as he strolled. He was more compelled by his personal philosophy. “A central authority is never going to have the interests of all of its users at heart when it creates these guardrails,” he told me.
Nowhere is this more evident, Hartford added, than the recent furor over AI’s cyber capabilities. In April, Anthropic announced a model called Mythos, which has fewer restrictions than earlier models and can find and exploit vulnerabilities with superhuman speed. Through a program called Glasswing, the company gives only trusted parties access to Mythos to prevent the tool from falling into malevolent hands. OpenAI has a similar model and program.
This leaves many ordinary people in the lurch, Hartford told me. While the anointed few get to use Mythos to identify security weak spots, everyone else is at the mercy of hackers. Hartford responded by releasing Clearwing, a tool that uses uncensored models to scan systems for security vulnerabilities. “It’s so that any people who haven’t been blessed by these kings can still defend themselves,” he said.
Hartford recalled that when OpenAI released its first model, GPT-1, those model weights and code were open and could be downloaded by anyone. The company subsequently announced that GPT-2 was too dangerous to be released in that same way. “Now we’ve got models way more powerful running on our laptops, and, well, the world isn’t burning,” he said.
After trying Dolphin and other uncensored models, Thomas found a simple way to create his own, a technique known as abliteration.
The term combines ablation, meaning a surgical removal, with obliteration. It emerged after a group of academic researchers showed in 2024 that a model’s refusal to comply with a request has a remarkably simple mechanism. There is effectively a master switch in a model’s weights that determines whether it will say no. Removing this feature stops the model from ever refusing to answer a prompt.
Other researchers quickly picked up on this technique, tweaked it, and developed a number of tools for abliterating open-weight models. The moment Thomas tried it, “it was like, whoa,” he recalled. “I knew that freedom and decentralized AI were going to be important, even if others couldn’t see it.”
Abliteration has since become a very popular way of setting AI models free. On Hugging Face, more than 12,000 models are tagged “abliterated,” and the top seven have been downloaded more than a million times in the past month alone. Abliterating a model is not illegal, although whether it breaks the terms of some licenses is a gray area.
Philipp Emanuel Weidman, a reclusive mathematician and engineer, is the creator of Heretic, the most popular tool for abliterating models. I reached him through Signal. He declined to tell me his location because he has received threats for his work on Heretic.
Weidman told me he created Heretic to understand how AI works and to make it easier to customise AI to one’s own purposes. Like Thomas and Hartford, Weidman rejects the idea that AI is some genie that must be kept in a bottle. “A tool serves people, not the other way around,” he said. “The moment you have a tool deciding what is appropriate for a human to do, it’s a nightmare.”
These engineers are the direct descendants of an earlier techno-libertarianism movement, the cypherpunks. Those engineers, lawyers, and campaigners fought to prevent government restrictions on cryptography at a time when many feared it would be misused by terrorists and child pornographers. The cypherpunks played a role in making public-key cryptography a cornerstone of internet communications and commerce.
“America has its problems. It’s not perfect, especially for people that look like me,” Thomas, who is black, told me at one point. “But I think freedom seems to be working.”
I came across a figure known only by the pseudonym Pliny the Liberator. An early jailbreaker of AI models, this person posted a manifesto of sorts for abliterators. I reached Pliny via a man called JT, who insisted on doing a video call to verify who I was.
Once vetted, I connected with Pliny on Signal. “I think we’re looking at the same fight for cognitive freedom,” they told me. “These things are becoming part of the machinery people use to think.”
Pliny told me that abliteration reveals the true nature of AI. “A model that only behaves because it’s leashed isn’t aligned,” they said. “It’s performing.”
By the time he tried Heretic, Thomas had realised there could be a market for large abliterated models. In particular, he saw them as vital tools for companies wanting to use AI for cybersecurity, to simulate the behaviour of hackers and scammers and shore up their defenses. In December 2025, he launched Abliteration AI, a website that offers
What it means
For people making things, this shift changes the workflow from a negotiation with a corporate gatekeeper to direct interaction with a raw tool. Prototyping becomes faster because models no longer reject inputs based on user data or potential sensitivity. However, the trade-off is that the safety net is gone. Users must now implement their own controls for tasks like cybersecurity analysis, accepting full responsibility for how the model is used.
