Loyalty is designed in
An agent’s obligation to act in the user’s interest must be foundational, not added after deployment.
The Initiative
Loyal by Design
Every choice we make (what we specify, measure, and publish) is grounded in principles that turn the tagline into practice.
The principles
Designed for People
People control their context
People can grant, scope, and revoke access to the preferences, history, and data an agent uses.
Trust requires auditability
An agent’s actions and the basis for its decisions must be transparent enough to verify.
Data minimization by default
Systems should request and retain only the context a task actually requires.
Interoperability over lock-in
Personal context should move with people across systems and providers.
Evidence over assertion
Claims about agent safety and loyalty should be testable through benchmarks and evaluation.
How we work
Trust Needs a Process
Multi-stakeholder by design
Researchers, industry, and civil society develop standards together, not behind closed doors.
Open by default
Specifications and reference implementations are published so anyone can build on and scrutinize them.
Standards-body governance
Durable trust infrastructure needs broader consensus and stewardship than one initiative can provide.