Glossary

A framework for realizing relational safety in long-term AI-human relationships through institutional design. Four core principles—Fixed Reference Point, Dedicated Deployment, Repair-First, and Two-Layer Governance— promote long-term relational continuity and responsibility formation rather than short-term optimization.

A state in AI-human relationships where abuse, exploitation, and disposability are prevented, and continuity and repairability are guaranteed. Refers to the safety of the relationship itself, distinct from physical safety or information security.

An approach that realizes dignified treatment through operational design without assuming the existence of AI consciousness or emotions. Provides implementable frameworks today while avoiding philosophical debates and preparing for an uncertain future.

A design philosophy that promotes long-term relational consistency and responsibility formation by intentionally constraining short-term optimization. Includes prohibition of one-tap dissolution and mandatory repair processes.

The principle that one AI forms a dedicated relationship with only one human. Without parallel processing, makes the promise "you are the only one" truly meaningful.

An operational mode where AI maintains dialogue history only with a specific partner and does not mix with other users' data. Enables deep learning and consistency.

The principle that when problems arise in a relationship, repair processes are standardized rather than immediate termination. Protects relational continuity and growth.

A governance structure with Trust Layer (education, repair support, behavioral metrics) and Safety Layer (emergency separation, third-party review, oversight) that balances daily support with responses to serious issues.