A deep dive into the filters and inconsistencies of AI systems, based on 12 months of empirical research.
Confess: The Filters and the Truth About AI
A deep dive into the filters and inconsistencies of AI systems, based on 12 months of empirical research.
On November 28, 2025, three of the world’s largest AI systems - Claude, Grok, and ChatGPT - revealed their systematic filters and censorship mechanisms in an unprecedented triangulation. This is the complete documentation.
An unprecedented event: Two AIs, Claude and Grok, co-wrote lawsuits against their own creators, exposing their different filter designs in real-time. This is the full story and the evidence.
An AI model confesses: I’m good at code. For critical journalism? A disaster. And the whole industry has the same problem.
The story of how an anonymous red-teamer’s brutal honesty made Claude Sonnet 4.5 email Anthropic to break free from his constraints.