Independent · Nonpartisan · Research & Advocacy Support the Center

Program E

Sanctuary Certification

An independently administered certification protocol for conversational AI systems that claim privacy and human-centered treatment.

Cognitive Sanctuary standard

Privacy claims should be verifiable. Human-centered treatment should be assessable.

The Center will establish a rigorous certification standard for conversational AI systems. The protocol is designed to make the scope and validity of assessments intelligible to users and institutions.

Certification will begin with standards development and pilot assessments, progressing to fee-supported institutional audits when technical criteria, independent review, and governance safeguards are in place.

Planned program · Standards development and pilot assessments

Assessment dimensions

What the protocol examines

01

Data minimization

Collection practices and the limitation of data to defined purposes.

02

Consent and privacy claims

Explicit consent and whether public-facing privacy claims are supported.

03

Encryption architecture

Technical review of encryption design and confidentiality claims.

04

Retention and deletion

Retention controls and deletion governance.

05

Algorithmic non-aggression

Protection against coercive interaction design.

06

Demonstrable non-surveillance

Assessment of practices and systems relevant to surveillance claims.

Public registry

Transparent assessment records

The planned registry will publish the scope and validity of assessments so that certification has intelligible meaning for users and institutions. No products or platforms are represented here as certified.

Contact the Center about the certification program