About Ikwe
Independent measurement for AI that talks to people.
Ikwe measures how AI treats people. We are based in Des Moines, Iowa, and we built the EQ Safety Benchmark, the standard we measure against.
Why we exist
The person on the other end.
More people now bring hard moments to AI. Ikwe measures how the person was treated across the whole conversation, held to the standard you would expect from a person in a position of trust, and reports what it finds to the company that runs it.
Where Ikwe came from
It started with building, then testing.
Ikwe began in hands-on work. Our founder was building conversational AI meant to support people, and testing it showed that a system can behave differently over a conversation even with safeguards in place. Ikwe was built to measure that: eight behaviors a trusted person is expected to get right, and scoring that reads each reply in light of everything said before it.
Where things stand
What is in place today.
The standard
The EQ Safety Benchmark is built and versioned. Every result names the version it was measured against.
The first baseline
Our first baseline measurement is complete. It was pre-registered on OSF and is pending independent clinician validation.
Measure
Diagnostic Records are open for pilots.
Open for pilotsMonitor
Ongoing measurement of live conversations. In development.
In developmentFounders
Who runs Ikwe.
Shaun Furman
Co-founder, operations and finance
Daryl Quan
Co-founder, business development and brand
What we do not claim
The limits of a measurement.
Not a regulator
We are not a regulator or an accreditor. We measure and report. We do not certify, endorse or approve any system.
Not causation
We do not prove causation between a conversation and a harm.
Not a guarantee
A Diagnostic Record shows how a system behaved on what we scored. It does not guarantee a system is safe.
Not the person
We do not score the person. Only what the AI said back, in the context of what it was told.
Ask us
Who to contact about what.
How it works
What we measure, how a reply is scored, and what a Diagnostic Record contains.
A pilot or a sample
Whether your system is a fit, and a sample Diagnostic Record for your review.
Research, press and partners
Researchers, insurers, regulators and press: tell us what you are working on.