Research

What we're learning, and where we stand.

Chrysalis is built on a point of view about AI, data ownership, and consent. This is where we publish the research and findings behind it, starting with the questions trust-based businesses are facing right now.

White paper 01

Helpful and Exposed.

The more helpful a tool is, the more of a person it needs.

By Brittany Martin · Builder and graduate researcher in data ownership, consent, and AI ethics

AI only gets more useful the more of yourself you give it. For anyone doing trust-based work, coaches, clinicians, creators, that's the exposure: the context that makes AI helpful is the same context that puts a person at risk. This paper lays out what practitioners are running into, what the research establishes, and a simple framework for deciding what to trust.

What's inside
  • 01What practitioners are encountering
  • 02What the research establishes
  • 03A practitioner decision framework
  • 04How Chrysalis applies the framework
Our position

Train memory, not models.

The more AI knows you, the more that knowledge should belong to you. Our research keeps landing on the same place: personalization and privacy are not opposites, but you only get both when the person owns their own context and decides what's shared. It's the principle underneath everything we build.

More research

More on the way.

We publish as the work matures. A few of the pieces we're developing next:

In progress

Consent as a system

Why consent has to be a live control surface, not a one-time checkbox, and what that looks like in practice.

In progress

The context layer

Where a person's context meets an expert's method, and why portability changes the economics of trust-based work.

In progress

Invisible illness & patient-owned data

Findings from our HHS Invisible Illness work on making the invisible visible without surveilling the patient.

Start with the first one

Get “Helpful and Exposed.”

Our founding research on AI, data ownership, and consent. We'll send it to your inbox.