Devon Thomas launched Abliteration AI to provide open-weight models without corporate safety guardrails, according to a Wired report.reported The company makes these uncensored versions available to anyone, challenging the industry norm that restricts model outputs.A new company lets users run AI models without the usual safety blocks.Abliteration AI distributes open-weight models with alignment and safety layers stripped.
Evidence: reported
Tear Down Those Guardrails! Meet the Renegades Freeing Chatbots From Corporate Enslaversaifleet-wired
The figures
- 36: Thomas’s age at the time of the report.
- 2020: Year Thomas joined Apple as an engineer.
- November 2025: Start date for Thomas’s AI rebellion and Abliteration AI venture.
Behind the figures
Wired reported these details through direct correspondence with Thomas via text and phone.reported The outlet describes Thomas as polite and kind, contrasting his demeanor with his subversive business model.
What it means for your stack
Developers using major hosted models face automatic refusals for inputs containing user information or cybersecurity data. Abliteration AI offers an alternative where models do not reject such queries, potentially streamlining prototyping workflows that involve sensitive log analysis.
Building on this
Engineers can build tools that analyze cyber attack logs without triggering refusal mechanisms. They can also prototype features that identify likely paying users, tasks that hosted models often block. The documented capability is unrestricted output for inputs deemed sensitive by major labs.
Why it matters
- Developers can bypass automatic refusals when analyzing cyber attack logs for defense purposes.
- Prototyping workflows involving user data no longer require workarounds for hosted model restrictions.
- Open-weight models are now available with safeguards removed, altering the risk landscape for deployment.
- Access to uncensored models is democratized, moving beyond paid tiers offered by major AI labs.
Topics: open-weight models, AI safety, Abliteration AI, Devon Thomas, cybersecurity
Updated
Then / now / next
Hosted AI models frequently refused legitimate cybersecurity and user data analysis requests.
Abliteration AI offers open-weight models with these safety guardrails removed.
It remains unclear how widely developers will adopt uncensored models for production use.
Story timeline
- Then Hosted AI models frequently refused legitimate cybersecurity and user data analysis requests.
- 10 Nov 2026 Tear Down Those Guardrails! Meet the Renegades Freeing Chatbots From Corporate Enslavers
- Now Abliteration AI offers open-weight models with these safety guardrails removed.
- Next It remains unclear how widely developers will adopt uncensored models for production use.
Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Terms in this story
open-weight models
AI models whose weights are publicly available, allowing independent hosting.
guardrails
Safety mechanisms that prevent AI models from generating certain types of content.
refusals
Automatic rejections of user inputs by AI models based on safety policies.
Questions readers ask
What company removes AI guardrails?
Abliteration AI, founded by Devon Thomas, removes safeguards from open-weight models.
When did Thomas start this venture?
His AI rebellion began in November 2025.
Where did Thomas work when he started Abliteration AI?
He was an engineer at Apple when his AI rebellion began in November 2025.
Connected stories
Reading receipt
- Devon Thomas founded Abliteration AI to remove safeguards from powerful open-weight models.
- Thomas worked at Apple as an engineer when his AI rebellion began in November 2025.
- The service targets users blocked when analyzing cyber attack logs or user behavior data.
Still open: How will Abliteration AI handle legal liability for misuse of its models?
Still open: Will major AI labs respond with stricter licensing for open-weight models?