Honest Framing
Emori is a monitoring system. We say so honestly.
A voice-based companion that observes daily patterns, logs what it notices, and reports to caregivers is a monitoring system. Describing it as anything else would be dishonest — and the people it serves deserve honesty above all.
What distinguishes Emori is not the absence of monitoring. It is the governance of it. Who decides what is observed. Who sees what. Who owns the data. And what can never be done with it, regardless of who owns Emori in the future.
What Emori monitors: Daily conversation patterns, word-finding, emotional tone, medication confirmation, session length, and engagement signals. Emori observes behavior — not content. Lumi notices patterns, not words.
What Emori never monitors: The content of any conversation. Any words spoken. Any names mentioned. Any story told. The content of every memory, every conversation, belongs entirely to the person and their family.
Data Governance
What is collected, who sees it, and who owns it.
Every interaction with Lumi generates two simultaneous data streams. They are separated at the source — before any data moves anywhere.
The path from research data back to the individual does not exist — because it was never built. One encrypted lookup table connects anonymous research identifiers to real accounts. It exists solely to enable consent withdrawal. It never leaves Emori's most secure environment and is never shared with the Serraro Research Trust.
The family owns Stream A. Serraro LLC holds it under the governance framework. It is never sold. It is never used for advertising. It is never shared without explicit consent. If you leave Emori, your data leaves with you.
Serraro LLC cannot access the research dataset for its own commercial purposes. The Trust holds it independently. If Serraro LLC is acquired, the research data does not transfer to the acquirer. This is built into the legal architecture — not promised in a policy.
Consent Architecture
Layered, person-controlled, and never bundled.
The clinical insight layer has two separate consent tracks. A family may say yes to one and no to the other. They are never bundled, never pre-checked, and never designed to nudge toward yes.
Both consent moments are honest about uncertainty. Track Two in particular cannot promise outcomes. The language reflects what is actually known — and what is not.
A note on Track Two — transparency about our default. Research contribution is enrolled by default. When a family completes registration, anonymous pattern data is contributed to the Serraro Research Trust unless they actively decline. We made this choice deliberately — and we say so openly.
Our reasoning: the populations Emori serves are the ones most absent from dementia research. Every family that participates brings us closer to care that is more personal, more effective, and more human for everyone who comes after. We believe most families want to be part of something bigger than themselves — and we designed the default to reflect that belief.
What makes this ethical is not the absence of a default. It is the honesty about it. The opt-out is a single checkbox, clearly labeled, never buried. Withdrawal is available at any time through the caregiver portal. Data is removed from the active research dataset within 30 days of withdrawal. The choice is real, reversible, and explained in plain language before it is made.
Track One — personal clinical report — remains a true opt-in. No default. No pressure. Separate decision entirely.
The Serraro Research Trust
An independent nonprofit that Emori cannot control.
Emori operates as two legally distinct entities. Serraro LLC builds and operates the product. The Serraro Research Trust — an independent nonprofit — holds and governs the anonymized research dataset.
The Trust is governed by an independent board that does not include Emori executives. Its board includes clinicians, ethicists, community representatives, and family advocates. Serraro LLC contributes data to the Serraro Research Trust under formal agreement — but does not control it.
Serraro LLC is structured to prevent us from exploiting this data even if we wanted to. The people who generated it deserved that protection built into the architecture — not just promised in a policy.
A new owner of Emori gets the product, the technology, the brand. They do not get the research data. It lives in the Serraro Research Trust. The Trust's charter governs it. This protection exists from the beginning — not added later when it is too late.
The Trust is committed in writing now — before the data exists, before the Serraro Research Trust is formally incorporated, before any research relationship is established. Phase 1 is the commitment. Phase 2 is the pilot, with consent boxes present and the research track honestly disclosed as not yet active. Phase 3 is formal establishment at scale.
The Research Gap
The communities Emori serves are the ones most absent from dementia research.
Most dementia research data comes from clinical settings — controlled, infrequent, anxiety-producing. It is conducted primarily in English. The populations it studies skew white, educated, and English-speaking. Spanish-speaking elders from Latin American backgrounds are almost entirely absent. Filipino, Japanese, Korean, and other immigrant communities are studied in fragments, if at all.
Naturalistic longitudinal data — daily observation at home, in a person's own language, over months and years — does not currently exist at any meaningful scale.
The research gap is not an abstraction. It means that the grandmother who immigrated from Sonora in 1981, who speaks Spanish and carries the specific cognitive patterns of her culture and her history, has almost never been studied. Emori can be the one that looks.
The community benefit requirement built into commercial licensing of Trust data is not charity. It is a structural correction to a historical pattern. The community that gave the data receives something back. That's not optional. It's the architecture.
AI Infrastructure
Infrastructure is not neutral. Ours was a deliberate ethical choice.
Lumi is powered by Claude, the AI model developed by Anthropic. This was not a technical default. It was a deliberate choice made for reasons inseparable from Emori's mission.
Anthropic was founded with a specific mandate: to develop AI that is safe, interpretable, and beneficial. For a product serving people with cognitive decline — among the most vulnerable users of any AI system — infrastructure built by people who take that responsibility seriously is not optional. It is the baseline.
— Emori onboarding disclosure, plain language version
Emori discloses its AI infrastructure to families. Not in fine print. In plain language, at onboarding, as part of the trust conversation. Families who bring Emori into their home deserve to know what it is built on and why those choices were made.