Last week, Consumer Reports and the Stanford Digital Economy Lab at the Human-Centered AI Institute co-hosted a workshop exploring the future of consumer-authorized AI agents. You can read more about our vision here.
The workshop brought together experts from Google, OpenAI, Microsoft, Visa, Salesforce, Genesys, and many other innovative companies and research organizations. Participants gathered to discuss the technical and legal foundations necessary for agentic systems that act in the best interests of consumers. Here are some key takeaways and insights from the event:
Need for Reference Patterns & Best Practices
One of the clearest points of agreement was the need to rapidly establish concrete, implementable best practices for consumer-authorized AI agents. While our initial focus is on authentication and authorization, with OAuth-based solutions viewed as practical for the short term—there’s recognition that we’ll ultimately need more robust solutions. The group emphasized balancing user-friendly interactions with strict protections around privacy, security, and user consent.
Big Challenges Ahead
We identified several thorny challenges, which we’ll begin tackling in the months ahead:
- Authentication & Consent: How will the consumer authorize the agent to act on their behalf, and how will that agent authenticate with companies? How do we ensure robust yet frictionless permissions without causing “permission fatigue”? How might we deliver simple ways for consumers to dynamically manage and revoke permissions?
- Liability & Accountability: Who is responsible when an agent messes up? As agents become more autonomous, establishing norms for liability, as well as transparent and traceable interactions, will be critical.
- Consumer Control & Privacy: Participants agreed strongly that safeguarding users from surveillance, dynamic pricing, and unauthorized actions is vital. Consumers need to have granular control, clear explanations of agent behaviors, and the ability to easily revoke or adjust permissions. The goal should be an agentic commerce ecosystem that consumers actively manage rather than one that manages them.
- Consumer Choice: Enabling interoperability across different interactions will be critical, whether those interactions are agent-to-agent (A2A), agent-to-consumer (A2C), or consumer-to-agent (C2A). Consumers need to be able to choose the right agent for them, and switch between agents seamlessly.
- Enterprise & Market Dynamics: Businesses will adopt these standards faster if they’re shown to enhance customer relationships and loyalty programs. The marketing landscape could shift dramatically as consumer-agent interactions become more common.
What’s Next?
Building on the momentum of this initial session, we’re quickly moving into action. Over the coming weeks, we’ll identify some consumer scenarios to rapidly prototype and test in sandbox environments. These early proofs of concept will set the foundation for evolving robust best practices, guided by real-world data and immediate user feedback.
We’ve got a lot of work ahead of us. We’re actively seeking input from forward-thinking organizations and stakeholders interested in defining and selecting use cases, building out prototypes, and shaping the next generation of authenticated delegation standards. Early partners will have a unique opportunity to influence the future of consumer-agent interactions and position themselves at the forefront of this transformative shift.
We’d like to extend a huge thanks to our partners at Stanford for helping us create such an energizing, productive session.
The remit for us is clear – empowering consumers with trusted agents is important for people, for the marketplace and for a future we want to live in. If you’re passionate about working with us to shape this rapidly evolving space, please reach out via InnovationLab@cr.consumer.org.
