Abliteration.ai (@abliteration

Abliteration.ai

118 posts

Abliteration.ai

@abliteration_ai

LLMs for high risk industries like security, defense, and trust & safety. Compatible API & training data generation.

Palo Alto, CA

abliteration.ai

Joined November 2025

Following

Followers

Pinned
Abliteration.ai
@abliteration_ai
May 15
Most teams do not need an “anything goes” model. They need an LLM that can generate the edge cases mainstream models refuse: synthetic data, eval cases, adversarial examples, and domain-specific test sets. That is what we are building at Abliteration.ai.
Unrestricted. Not ungoverned. — Developer-controlled LLM API
From abliteration.ai
721
Abliteration.ai
@abliteration_ai
Jun 11
Replying to @abliteration_ai
In the end, the centralization of control to the big labs is unsustainable and the future will prove this.
38
Abliteration.ai
@abliteration_ai
Jun 11
Replying to @abliteration_ai
We believe that there are many legitimate models for less-restrictive models and we will continue to push for them to be available.
46
Abliteration.ai
@abliteration_ai
Jun 11
Replying to @abliteration_ai
original article:
It blocked us at 'hello!' Anthropic Fable 5 refusing innocuous prompts
From theregister.com
67
Abliteration.ai
@abliteration_ai
Jun 11
We love the update
240
Abliteration.ai
@abliteration_ai
Jun 11
Quoted in @TheRegister today on why Anthropic’s overzealous guardrails are backfiring. “Anthropic’s making a big bet on their brand that people will trust their brand so much they’ll just deal with [refusals]. But in the long term, people are not just going to accept these
95
Abliteration.ai reposted
Founder Engineer
@founderengineer
Jun 10
Why we built @abliteration_ai
Joseph Thacker
@rez0__
Jun 9
It’s hilarious that they made a huge deal about the cyber capabilities for months and then when they rolled it out, they’ve blocked the actual utility of the model by prohibiting cyber use 🤣 And yes this includes trusted testers. Like, what was the point in even releasing it?
289
Abliteration.ai reposted
Founder Engineer
@founderengineer
Jun 9
Exactly why we built @abliteration_ai
Zack Korman
@ZackKorman
Jun 9
Mythos is amazing.
349
Abliteration.ai
@abliteration_ai
Jun 8
We need to model bad behavior before it happens in the real world.
Founder Engineer
@founderengineer
Jun 8
Article
Model Bad Behavior
The people responsible for finding failures should find them before attackers or abusive users. Security teams, AI evaluators, and trust and safety teams need to understand what can go wrong. If the...
664
Abliteration.ai reposted
Founder Engineer
@founderengineer
Jun 5
When my cofounder started working on integrating web search into @abliteration_ai ‘s data generation/chat UIs and APIs I was skeptical of how good we could make it but figured it would be worth a shot. It turned out GREAT especially for generating modern, state of the art cyber,
GIF
247
Abliteration.ai reposted
Founder Engineer
@founderengineer
Jun 2
We at @abliteration_ai have been telling anyone who would listen that the good guys (security professionals) need the same (or better) AI tools as the bad guys. Today, the White House has said two different things. One, labs should have models reviewed before release; two, we
Promoting Advanced Artificial Intelligence Innovation and Security
From whitehouse.gov
159
Abliteration.ai reposted
Founder Engineer
@founderengineer
May 30
People keep asking what the point of abliterated models is. Red Hat just published a good example of why you would want one for red teaming and infrastructure testing. We host them for you at @abliteration_ai and you just use the OpenAI and anthropic-compatible apis
Testing infrastructure red teaming with abliterated models | Red Hat Developer
From developers.redhat.com
128
Abliteration.ai reposted
Founder Engineer
@founderengineer
May 29
Testing out @KiteVideoHQ for a video for @abliteration_ai ‘s training data generation feature for security/trust & safety use cases
00:00
546