01
Can we build a lie detector for AI?
Develop rigorous white-box methods to detect deception, hidden goals, censorship, hallucination, reward hacking, and failures missed by ordinary evaluation.
Auditing AIICINA · International Consortium for Interpretable AI
A new science for understanding machine intelligence.
ICINA brings researchers and industry together to build evidence-based methods, shared infrastructure, and a scientific commons for interpretable AI.
01
As AI capabilities rise, human agency must grow.
Why ICINA
Understanding is the foundation of responsibility.
Today’s neural networks are trained, not programmed. Their internal mechanisms emerge from data, leaving society with the challenge of reverse-engineering the learned programs that underlie their behavior.
Scientific priorities
Three questions that cannot wait.
We focus the field on concrete goals that connect understanding with public responsibility.
02
What is thinking?
Use fully observable silicon brains to investigate concepts, abstraction, belief, planning, learning, agency, and the causal structure of reasoning.
Understanding intelligence03
How can AI amplify human agency?
Translate discoveries at the edge of machine knowledge into human insight, better judgment, and more capable human–AI collaboration.
Empowering peopleOur approach
Shared investments no single laboratory can sustain alone.
ICINA will convene researchers, support talent, and build open scientific infrastructure across institutions and sectors.
Talent
Fellowships, visiting researchers, and protected service rotations that bring people together across institutions.
Shared infrastructure
Measurement, Models, Methods, Metrics, and Machines: the five foundations of reproducible interpretability research.
Convening
Research gatherings and public technical discussions centered on human teaching, learning, and scientific consensus.
Globally networked
One consortium.
Three initial hubs.
BostonResearch universities
San Francisco Bay AreaFrontier laboratories and startups
LondonAcademic and AI-safety community
Founding consortium
Help establish a field built for shared evidence.
We are assembling the initial group of academic and industry members who will shape ICINA’s research programs, infrastructure, and standards.
Discuss participation