Loyal Agents

Stanford × Consumer Reports

AI Is Done Chatting.
It’s Making Decisions.

We’re defining the standards for agent behavior so that every decision your agent makes is optimized for your benefit, security, and privacy.

Fund This InitiativeJoin as ResearcherShape Policy

Definition

What Is a Loyal Agent?

A loyal agent is an AI agent that acts in the best interests of the person it serves.

  • Acts for you

    It puts your interests and stated preferences first.

  • Discloses incentives

    It tells you when it has a conflict of interest.

  • Protects your data

    It uses your information only within the limits you set.

  • Can be checked

    Its behavior can be tested and audited.

Loyal Agents is the initiative working to make this the norm. It is a research collaboration between the Stanford Digital Economy Lab and Consumer Reports that develops the legal frameworks, open standards and evaluation tools for AI agents that are loyal by design.

See what we’re building

The stakes

When AI Acts on Your Behalf, Whose Interests Come First?

The digital economy is moving beyond assistive chat to delegated action, where AI agents negotiate, transact, and manage intimate details of our lives.

Economic

The organizations that define the standards will shape market structure and value distribution for decades.

Civic

Agents loyal to platforms rather than people risk entrenching power asymmetries and undermining autonomy.

Trust

Transparent, auditable agents must prove they serve you, and disclose any incentives that could pull them away from your interests.

The trust gap

Two Possible Futures

With standards

Agents work for you.

Transparent, auditable systems that prove their loyalty to the people they serve.

Without standards

Incentives stay hidden.

Your shopping assistant negotiates prices that benefit affiliates while claiming to work for you.

See the difference in a real request →

How the standard develops

How Our Standards Mature

We’re a research and standards initiative. Our work moves through stages of increasing scrutiny and adoption.

  1. Developing with reference partners

    Build and test specifications against real implementations with partners who surface problems a document alone cannot reveal.

  2. Widening review

    Use lessons from early implementations to shape the next version for a larger, hand-picked group of reviewers.

  3. Public circulation

    Circulate specifications publicly after testing and review by multiple independent implementers.

  4. Standards-body governance

    Bring mature work into an established body such as IEEE or W3C for broader consensus and stewardship.

Explore the Workstreams

Take part

Help Shape the Standards for AI Agents

Loyal Agents works with funders, researchers, policy leaders and early adopters. Find the path that fits you.

Funders & Founding Partners

Shape the future of the digital economy with early visibility into the standards that will guide how AI agents act for people.

Explore funding partnerships

Research Collaborators

Drive state-of-the-art inquiry and help define foundational protocols major platforms can use.

Explore research opportunities

Policy Leaders

Bridge fiduciary principles and high-performance systems, shaping the rules for AI agents that act on people’s behalf.

Get involved in policy and standards

Early Adopters

Put our standards and research to work early, starting with the Human Context Protocol, and collaborate on memory and AI safety research.

Become an early adopter

Momentum & impact

A Growing Coalition

Participants listening to a presentation at the Portable Memory + HCP workshop.
Portable Memory + HCP: Global Portable Memory Workshop, Stanford University, October 2026
20+Researchers & Collaborators
4Active Workstreams
11+Partner Organizations

Latest

Portable Memory + HCP: Global Portable Memory Workshop

Participants of the Portable Memory + HCP workshop at Stanford University, October 2026.

October 2026

Loyal Agents and the Gates Foundation convened researchers, builders and policymakers at Stanford University to work on how a person's agent memory can move with them across services.

Read more in Insights & Updates →