What we believe
Safety and capability are not opposites.
The assumption that we must trade one for the other is the most dangerous idea in AI development. Every capability advance we make is paired with corresponding safety research. Not as constraint — as architecture.
Transparency is a safety mechanism.
We publish our research, our evaluations, our mistakes. Opacity in frontier AI development is not caution — it is the absence of accountability. We choose accountability.
The alignment problem is solvable.
Not easily, not quickly, but tractably. The scientific evidence points toward solutions. We are assembling the team and the methodology to find them within this decade.
We think in centuries, not quarters.
Every model we ship, every paper we publish, every policy position we take is made with the understanding that these decisions will shape the trajectory of intelligence itself.
The researchers.
4 core team membersDr. Sarah Chen
Chief Executive Officer
Former lead researcher at DeepMind. PhD in Computer Science from MIT. Pioneer in constitutional AI and reward modeling. Believes the alignment problem is solvable within this decade.
Dr. Marcus Webb
Chief Research Officer
Former co-lead of OpenAI's safety team. PhD in Mathematics from Cambridge. Specializes in scaling laws, interpretability, and theoretical foundations of deep learning. 47 peer-reviewed publications.
Dr. Aisha Patel
Head of Interpretability
PhD in Cognitive Science from Stanford. Former researcher at Redwood Research. Leads AXIOM's mechanistic interpretability program, working to understand what frontier models actually compute.
Dr. Elena Vasquez
Head of Safety Evaluations
Former AI safety researcher at the UK AI Safety Institute. PhD in Statistics from Oxford. Leads the red-teaming and evaluation programs that gate every model release.